Anti-saturation and decoupling active-disturbance-rejection control method for afterburning type oxyhydrogen rocket engine

Abstract

The invention belongs to the field of afterburning type oxyhydrogen rocket engine control, and provides an anti-saturation and decoupling active disturbance rejection control method for an afterburning type oxyhydrogen rocket engine. The active-disturbance-rejection controller is used for solving the problems that parameter identification of a rocket engine is difficult, the chamber pressure mixing ratio is coupled, and the amplitude of an actuator is saturated. In order to solve the three problems, the method comprises the following steps: firstly, acquiring a control parameter matrix and an unmodeled dynamic state through a system identification method; and secondly, based on an identification result, rapid setting is carried out, an identification result of an unmodeled dynamic state is introduced into a controller, and a decoupling controller is designed. Thirdly, a control signal difference value before and after an actuator amplitude saturation module is introduced into an expansion state observer, and an anti-saturation and decoupling controller is designed; finally, simulation verification is carried out, and the result proves that the method can be rapidly set, and mixing ratio fluctuation caused by coupling during chamber pressure adjustment and chamber pressure and mixing ratio oscillation caused by actuator amplitude saturation are effectively reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of liquid rocket engine control, and particularly relates to an anti-saturation decoupling active disturbance rejection control method for a staged combustion hydrogen-oxygen rocket engine. BACKGROUND

[0002] With the increasing space exploration activities of human beings, higher requirements are put forward for the control system of variable thrust rocket engines. The staged combustion hydrogen-oxygen rocket engine is a typical multiple input multiple output (MIMO) system, and the most concerned control objects are the combustion chamber pressure (related to the engine thrust) and the mixture ratio (the mass ratio of oxidizer to fuel). The normal operation of the engine cannot be achieved without a reliable chamber pressure and mixture ratio control system. Due to the complex flight environment of the variable thrust liquid rocket engine, and the existence of component damage, changes in its own system parameters, chamber pressure and mixture ratio coupling, etc., the thrust and mixture ratio control system needs to have strong robustness. The active disturbance rejection controller (ADRC) is suitable for the design of the thrust and mixture ratio control system of the engine due to its low dependence on the model, strong dynamic disturbance compensation and decoupling ability, etc.

[0003] The active disturbance rejection controller has strong robustness and anti-interference ability, and its control effect depends on the reasonable setting of the control parameters, and the error between the estimated value and the actual value of the parameters should not exceed 30%. In order to ensure the control performance, the parameter matrix should be set to be close to the actual control matrix , and the trial-and-error method is often used to complete the setting in engineering. However, the control parameters of the rocket engine are unknown, and the matrix contains multiple variables, so the trial-and-error setting is high in cost and low in efficiency, and therefore the primary task of realizing the application of ADRC in the chamber pressure and mixture ratio control is to solve the setting of the matrix The decoupling ability of ADRC is derived from the real-time estimation and compensation of the coupling dynamics as part of the total disturbance, and its effect depends on the dynamic tracking performance of the Expansion State Observer (ESO). Increasing the bandwidth of the observer can enhance the tracking ability, but it will also amplify high-frequency noise, leading to output fluctuations. Therefore, how to achieve fast and accurate tracking of the disturbance by the ESO at a lower bandwidth and suppress the coupling effect has become a key difficulty in the design of the engine self-disturbance decoupling controller. In addition, model uncertainty and external disturbances in the actual system jointly affect the system performance, and ADRC unifies them as the total disturbance and estimates and compensates them through ESO to improve robustness. However, the physical structure of the actuator limits the amplitude of the control input, which weakens the disturbance compensation ability. Therefore, how to improve the ADRC to cope with actuator amplitude saturation has also become another difficulty in practical application.

[0004] In view of the above analysis, the application provides a thrust and mixing ratio closed-loop control design method for a staged combustion hydrogen-oxygen rocket engine. SUMMARY

[0005] In view of the above problems, the application provides an anti-saturation decoupling active disturbance rejection control method for a staged combustion hydrogen-oxygen rocket engine.

[0006] In order to achieve the above purpose, the technical scheme adopted by the application is:

[0007] An anti-saturation decoupling active disturbance rejection control method for a staged combustion hydrogen-oxygen rocket engine, the steps are as follows:

[0008] 1. Obtain the control parameter matrix and unmodeled dynamics of the engine through a system identification method

[0009] For the design of active disturbance rejection controller (ADRC) of the restartable liquid hydrogen / liquid oxygen rocket engine, the control parameters need to be determined first. The aerothermodynamic model of the restartable liquid hydrogen / liquid oxygen rocket engine can accurately reflect the dynamic characteristics of the engine. However, due to the complex structure of the engine and the involvement of various complex flow phenomena and multi-physical field coupling, it is difficult to establish an accurate engine mechanism model. The input and output of the engine contain its dynamic characteristics. In the case of known input and output, the engine model can be established by system identification method to obtain its control parameters. First, collect the historical operation data or test data of the engine under various working conditions, including control input (valve opening) and system output (chamber pressure, mixture ratio). Select a suitable discrete model to approximate the dynamic characteristics of the engine. Since the restartable liquid hydrogen / liquid oxygen rocket engine is a nonlinear model, the NARMAX (Non-linear Auto Regressive Moving Average with Exogenous Inputs) model is an input-output description of nonlinear models, which can efficiently capture the nonlinear, coupled and noise characteristics of the system. Therefore, the NARMAX model is selected as the discrete representation of the restartable liquid hydrogen / liquid oxygen rocket engine. To construct the NARMAX model, the appropriate structure term needs to be selected first. After the structure term is determined, the model structure term detection and parameter estimation are carried out by using appropriate algorithms, such as least squares method, gradient descent method, genetic algorithm, etc. Finally, a NARMAX model that can accurately reflect the dynamic characteristics of the engine is obtained. Based on this identification model, the control parameter matrix of the system near a specific working point can be obtained. By comparing the current output value with the product of the control parameter matrix and the current input value, the unknown dynamics at the current time can be obtained , which contains all the effects of coupling, non-coupling dynamics and external disturbances, providing a clear compensation target for the ADRC.

[0010] 2. Decoupled ADRC design

[0011] The decoupled ADRC (DADRC) is designed using the identification results of Step 1. The ADRC usually consists of three parts: tracking differentiator (TD), state error feedback control law (SEF), and ESO. The focus is on the design of ESO and SEF. First, based on the control gain coefficient matrix tuned . Since the performance of ADRC is better when is close to , set ​ or fine-tune nearby. Second, tune the ESO parameters and introduce unknown dynamics. ESO is the core of ADRC, which expands the internal and external uncertainties of the system model (i.e., "total disturbance") into a new state variable for observation. The numerical value of the system "unknown dynamics" calculated in the identification process is introduced into the ESO as prior knowledge to enhance the ESO's ability to observe the total disturbance, and a decoupled expansion state observer (DESO) is constructed to estimate the true value of the total disturbance more quickly and accurately . On this basis, in dynamic compensation, the total disturbance observed by DESO, the designed state feedback, and the identification value are used to generate the final control quantity together: , where is the output of SEF. In this way, the controller can actively compensate for disturbances and coupling effects on system output before they occur, greatly improving the robustness and dynamic performance of the system.

[0012] 3. Anti-windup decoupling active disturbance rejection controller design

[0013] To suppress the integral saturation phenomenon caused by actuator saturation, an anti-windup decoupling active disturbance rejection controller (ADADRC) based on the Anti-windup idea is designed on the basis of the DADRC designed in Step 2. The design idea is as follows: when the control command does not exceed the actuator limit , the system works as originally designed; when u reaches saturation, the compensation mechanism is activated. The compensation mechanism is as follows: the difference between the calculated control quantity and the saturated control quantity actually output by the actuator is monitored in real time. This difference is fed back to the ESO through an anti-windup compensation gain matrix . This feedback signal informs the observer in advance that the current control command has not been fully executed, and the actual state of the system will deviate from the observed value, causing the ESO to adjust its internal state in advance to make its observation value closer to the true dynamics of the system after saturation, effectively suppressing the excessive integral quantity accumulated due to saturation and preventing the controller from "losing control". After the control quantity exits saturation, the system can smoothly and without overshoot return to normal control mode.

[0014] 4. Simulation verification

[0015] The simulation result shows that the method can quickly set the control parameter matrix, effectively realizes decoupling and restrains the influence of saturation limitation.

[0016] The beneficial effects of the application are: 1. For the MIMO system, the system identifies the control coefficient matrix, and quickly sets based on the identification result , avoids the cost of trial and error method. 2. Through real-time calculation and compensation of coupling dynamics, the ESO can efficiently track the disturbance caused by coupling and suppress the influence of coupling. 3. By introducing the difference value of the control signal before and after saturation, the output oscillation caused by saturation limitation is suppressed. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is a structure diagram of a restartable hydrogen-oxygen rocket engine.

[0018] Figure 2 It is a chamber pressure and mixture ratio tracking error under different parameter matrices .

[0019] Figure 3 It is a chamber pressure and mixture ratio observation error under different parameter matrices .

[0020] Figure 4 It is an ADRC and DADRC chamber pressure comparison curve.

[0021] Figure 5 It is an ADRC and DADRC mixture ratio comparison curve.

[0022] Figure 6 It is an ESO and DESO total disturbance term observation value.

[0023] Figure 7 It is a chamber pressure and mixture ratio ESO and DESO estimation error curve.

[0024] Figure 8 It is Figure 7 a partial enlarged view.

[0025] Figure 9 It is a chamber pressure and mixture ratio estimation error curve of ADESO.

[0026] Figure 10 It is an ADADRC control input.

[0027] Figure 11 It is a DADRC, ADADRC chamber pressure and mixture ratio curve. DETAILED DESCRIPTION

[0028] The invention will be further described below with reference to the accompanying drawings, detailing the design of a self-disturbance rejection decoupling anti-saturation controller for a staged combustion hydrogen-oxygen rocket engine.

[0029] 1. Engine control parameters and unmodeled dynamic identification

[0030] Simplified diagram of engine system as follows Figure 1 As shown. For this engine, a control strategy is adopted: the pre-combustion chamber oxygen valve (POV) controls the thrust, and the thrust chamber oxygen valve (MOV) controls the air-fuel mixture ratio. The engine model is established as follows:

[0031] (1)

[0032] in, These represent engine chamber pressure and air-fuel mixture ratio, respectively. These represent the opening of POV and MOV, respectively. Indicates system control parameters, Indicates external disturbance. This indicates that the system is not dynamically modeled. Equation (1) is forward-differenced to obtain:

[0033] (2)

[0034] Equation (2) can be rewritten as:

[0035] (3)

[0036] in, Indicates the sampling period. Indicates time, , express The system dynamics of chamber pressure and mixing ratio at any given time are not modeled. and The identification problem is transformed into (3) and Identification.

[0037] Since the structural terms of the staged combustion hydrogen-oxygen rocket engine model are unknown, the NARMAX model, as an input-output description of a nonlinear system, can well characterize the dynamic characteristics of nonlinear systems. The NARMAX model is chosen to discretize the staged combustion hydrogen-oxygen rocket engine model, and then fitted using the NARMAX model to construct the corresponding linear parameter model:

[0038] (4)

[0039] in, This is for the output (chamber pressure, mixture ratio) of a staged combustion hydrogen-oxygen rocket engine. For model structure items, These are the parameters corresponding to the model structure terms. For model error terms, The number of structural items in the model;

[0040] The linear parameter model is a first-order system with input and output delays both being d. The model structure terms are set as follows:

[0041]

[0042] Based on the input and output data of the staged combustion hydrogen-oxygen rocket engine within a certain time period, the matrix form of the linear parameter model is obtained:

[0043] (5)

[0044] in, , , , , , ;

[0045] By tuning a suitable model coefficient matrix This makes the dynamics of model (5) approximate the dynamics of system (3) at the data point time. The RMGS algorithm (Recursive modified Gram-Schmidt orthogonalization method) is used to identify the control parameter matrix and the unmodeled dynamics of the staged combustion hydrogen-oxygen rocket engine model. Combined with ESS (Error Reduction Ratio), the model structure terms are selected and the orthogonalization order is adjusted. The unknown parameters are solved based on the identification results. Specifically, for Perform QR (orthogonal triangular decomposition) to initialize the model structure terms. Calculate the first The error reduction ratio (ESS) of each model structural term is used to select the structural term that contributes the most. The required prediction accuracy or iteration threshold is used as the stopping condition, and finally, the unknown parameters are solved by back substitution.

[0046] After obtaining the coefficients corresponding to the structural terms, The corresponding coefficient is denoted as ( ),but Identification value Represented as:

[0047] (6)

[0048] Recognition matrix for

[0049] (7)

[0050] Once the model input and output values ​​and B are known, the identification value of the unmodeled dynamics can be calculated. for

[0051] (8)

[0052] 2. Decoupling Controller Design

[0053] The unknown dynamics in the staged combustion hydrogen-oxygen rocket engine model can be obtained using the above method. Estimated information With unknown control gain matrix The estimated value Considering the performance of ADRC in terms of parameters near When it is better, Set as Or Minor adjustments were made nearby. right The estimate is accurate, By incorporating known information into the ESO, a low-bandwidth ESO can achieve rapid and accurate estimation of unknown dynamics caused by coupling, thereby suppressing the effects of coupling. When introducing information about... After obtaining the estimated information, according to formula (1), let ,definition For the total disturbance, System (1) can be equivalently represented in the following state-space form:

[0054] (9)

[0055] The designed IESO (Improved Extended State Observer) is as follows:

[0056] (10)

[0057] in int is the integer operator. Indicates time, Indicates the sampling period. It is the observer gain; They are The estimated value; Indicates the output of a staged combustion rocket engine; To measure noise, the observer gain is tuned according to the bandwidth method. , This refers to the observer gain that needs to be adjusted.

[0058] When the information of is unknown, the total disturbance in the model of the restartable hydrogen-oxygen rocket engine is Since the improved ESO contains partial estimation information of , the improved control law can be expressed as:

[0059] (11)

[0060] where , denote the controller gains of the chamber pressure loop and the mixture ratio loop, respectively. Equations (9)-(11) constitute the proposed decoupling active disturbance rejection controller for the restartable hydrogen-oxygen rocket engine; denotes the reference input; and

[0061] (12)

[0062] Since is set to be close to , equation (5) is approximated as

[0063] (13)

[0064] That is, the unknown term in the engine model is compensated, and since the coupling dynamics are contained in the unmodeled dynamics, the coupling is also eliminated, achieving decoupling.

[0065] 3. Anti-windup decoupling controller design

[0066] Considering the amplitude limitation, equation (1) is rewritten as:

[0067] (14)

[0068] Each component of

[0069] (15)

[0070] The control input is affected by actuator saturation, so the compensation effect of the total disturbance will be inhibited. The difference between the control amount u and the saturated control amount actually output by the actuator is fed back to the ESO through an anti-windup compensation gain matrix , and the designed anti-windup decoupling ESO is as follows:

[0071] (16)

[0072] where , , These are the anti-windup gains for the room pressure and mixing ratio loops, respectively. Equations (16) and (11) constitute the designed anti-saturation decoupling self-disturbance rejection controller.

[0073] 4. Simulation verification

[0074] Based on the designed anti-saturation decoupling active disturbance rejection controller, a model was built in the Simulink simulation environment, and its effectiveness was verified through room pressure and mixing ratio control. The controller gain and observer gain were tuned using the bandwidth method. (Controller bandwidth...) The larger the value, the faster the system response speed, but a larger value... This can lead to overshoot and oscillations in the system. On the other hand, the observer bandwidth... The larger the value, the higher the accuracy of the observer estimation, but a larger value... This will amplify the effects of noise. Based on the control effect, the observer bandwidth of the indoor pressure loop is tuned to... The controller bandwidth is set to The observer bandwidth of the mixing ratio loop is tuned to... The controller bandwidth is set to The observer and controller gains are... .

[0075] like Figure 2 , Figure 3 As shown, compared with the trial-and-error method, the system adjustment time is significantly shortened and the observation error is reduced after tuning using the control parameter matrix identification results, and no obvious oscillations are observed.

[0076] Figure 4 and Figure 5 The effects of ADRC and DADRC on indoor pressure and mixing ratio control are demonstrated. With similar indoor pressure regulation performance, the DADRC decoupling strategy results in smaller mixing ratio fluctuations and a faster return to the set value.

[0077] Figure 6 The total disturbance estimation curves for ESO and DESO are shown. In the ventricular pressure loop, since no identification values ​​for unmodeled dynamics are introduced, the estimation curves for ESO and IESO are basically the same. However, in the mixing ratio loop, the total disturbance estimate for DESO is close to zero, while the estimate for ESO is larger. This is because DESO introduces identification information for unmodeled dynamics, requiring less unknown dynamics to be estimated than ESO, thus DESO has better observation performance under the same bandwidth.

[0078] Figure 7 The observation errors of ESO and DESO regarding indoor pressure and mixing ratio were compared. Figure 8This is a magnified view of the observation error of the mixing ratio loop. Under the same observer gain, the observation errors of the chamber pressure loop are similar; however, due to the introduction of unmodeled dynamic identification information, the observation error of the mixing ratio loop (DESO) is significantly lower than that of the ESO.

[0079] Figure 9 The observation error of ADESO converged to 0. Figure 10 The system indicates that actuator saturation has occurred. Figure 11 The control effects of ADADRC and DADRC were compared, and the results showed that ADADRC can effectively reduce the output oscillation caused by actuator saturation.

Claims

1. A method for anti-saturation decoupled active disturbance rejection control of a staged combustion hydrogen-oxygen rocket engine, characterized in that, Firstly, the control parameter matrix and unmodeled dynamics of the engine are obtained by system identification method; secondly, the unmodeled dynamics are introduced into the controller based on the identification results, and the decoupling controller is designed; thirdly, the difference between the control signals before and after the actuator amplitude saturation module is introduced into the extended state observer, and the anti-saturation decoupling controller is designed; finally, the simulation verification is carried out.

2. The anti-windup decoupled active disturbance rejection control method for a restartable liquid hydrogen-oxygen rocket engine according to claim 1, wherein, The specific steps are as follows: Step 1, the control parameter matrix and unmodeled dynamics of the engine are obtained by system identification method The model of the engine is established as: (1) ； wherein, respectively denote the engine room pressure and the mixture ratio, respectively denote the opening of the POV and the MOV, denotes a system control parameter, denotes an external disturbance, denotes system unmodelled dynamics; equation (1) is forward-differenced, yielding: (2) ； Rewrite equation (2) as: (3) ； wherein, denotes the sampling period, denotes the time instant, , denotes the unmodelled dynamics of the system in terms of chamber pressure and mixing ratio; and the identification problem of (3) reduces to the identification of and . The NARMAX model is selected to discretize the model of the afterburning hydrogen-oxygen rocket engine, and the NARMAX model is fitted to construct the linear parameter model corresponding to the NARMAX model: (4) ； wherein, is the output of the staged combustion hydrogen-oxygen rocket engine, is the model structure term, is the parameter corresponding to the model structure term, is the model error term, is the number of model structure terms; The linear parameter model is a first-order system, and the input and output time delays are d. The model structure term is set as: ； According to the input and output data of the afterburning hydrogen-oxygen rocket engine in a certain period, the matrix form of the linear parameter model is obtained: (5) ； wherein , , , , , ; By tuning a suitable model coefficient matrix This makes the dynamics of model (5) approximate the dynamics of system (3) at the data point time; the RMGS algorithm is used to identify the control parameter matrix of the staged combustion hydrogen-oxygen rocket engine model and the unmodeled dynamics of the engine, and the error reduction ratio ESS is used to select the model structure terms and adjust the orthogonalization order, and the unknown parameters are solved based on the identification results; specifically, for Perform QR decomposition and initialize model structure terms. Calculate the first The error reduction ratio (ESS) of each model structural term is used to select the structural term that contributes the most based on the error reduction ratio. The required prediction accuracy or iteration threshold is used as the stopping condition, and finally the unknown parameters are solved by back substitution. After the coefficient corresponding to the structure term is derived, the structure term is added to the model The coefficient corresponding to the structure term is denoted as , Then The identified value of the structure term is denoted as ​ (6) ； the recognition matrix of is: (7) ； After the model input-output values and B are known, the identified value of the unmodeled dynamics can be calculated, and the identified value of the unmodeled dynamics To (8) ； Step 2, based on the identification results, the unmodeled dynamics are introduced into the controller, and the decoupling controller is designed The unknown dynamics in the staged combustion hydrogen-oxygen rocket engine model were obtained using the above method. Estimated information With unknown control gain matrix The estimated value Considering the performance of ADRC in parameters near Shi Youwei, will Set as Or Minor adjustments nearby; right The estimate is accurate, By incorporating known information into the ESO, a low-bandwidth ESO can achieve rapid and accurate estimation of unknown dynamics caused by coupling, thereby suppressing the impact of coupling; when introducing information about... After obtaining the estimated information, according to formula (1), let ,definition For the total disturbance, The system (1) is equivalently represented in the following state-space form: (9) ； The designed IESO is as follows: (10) ； wherein int is an integer operator, denotes time, denotes a sampling period; is an observer gain; are respectively estimated values of denotes the output of a restartable hydrogen-oxygen rocket engine is a measurement noise; the observer gain is set according to the bandwidth method as , is an observer gain to be adjusted; When the information of is unknown, the total disturbance in the model of the staged combustion hydrogen-oxygen rocket engine is Since the improved ESO contains partial estimation information of , the improved control law is represented as (11) ； in, , The controller gains of the chamber pressure loop and the mixture ratio loop are respectively represented; Equations (9)-(11) constitute the proposed decoupled self-disturbance rejection controller for the afterburning hydrogen-oxygen rocket engine; Indicates the reference input; substituting equation (11) into system (1) yields (12) ； Due to tuned to be close to , equation (5) is approximated as (13) ； That is, the unknown term in the engine model is compensated, and since the coupling dynamics is included in the unmodeled dynamics, it is also eliminated, realizing decoupling; Step 3, the difference between the control signals before and after the actuator amplitude saturation module is introduced into the extended state observer, and the anti-saturation decoupling controller is designed Considering the amplitude limitation, equation (1) is rewritten as: (14) ； Each component of the vector is represented as: (15) ； The control input is affected by actuator saturation, so the compensation effect of the total disturbance will be inhibited, and the difference between the control amount u and the saturation control amount actually output by the actuator is fed back to the ESO, and the anti-saturation decoupling ESO is designed as follows: by an anti-saturation compensation gain matrix , which is fed back to the ESO. (16) ； wherein , , are the anti-windup gains of the room pressure and the mixing back loop, respectively; equations (16), (11) constitute the designed anti-windup decoupled active disturbance rejection controller; Step 4, simulation verification.

Images