Engine nozzle actuator adaptive control method and device

Abstract

This invention discloses an adaptive control method for an engine exhaust nozzle actuator. The invention improves upon the traditional Smith predictor and combines it with a second-order linear LADRC controller to control the engine exhaust nozzle actuator. The improved Smith predictor includes a traditional Smith predictor and a first-order filter. The output y1 of the system's built-in model of the traditional Smith predictor after a time delay is compared with the output y1 of the controlled object. p The difference dy between the two values ​​is filtered by the first-order filter to compensate for model inaccuracies. The output y of the traditional Smith predictor's built-in model does not go through a time delay element. p The sum of the output signal of the first-order filter and the output signal of the first-order filter is used as the tuned system output y′ of the improved Smith predictor. This invention also discloses an adaptive control device for an engine exhaust nozzle actuator. This invention can maintain high control accuracy even when the exhaust nozzle actuator experiences various degradations such as gain degradation and delay degradation.

Description

Technical Field

[0001] This invention relates to a control method for an engine exhaust nozzle actuator, belonging to the field of aerospace engine control technology. Background Technology

[0002] To ensure high kinetic energy in aero-engine exhaust, the design performance of the tailpipe must be carefully calculated beforehand. Since the flight process of an aero-engine encompasses its entire flight envelope and its operation is extremely complex, the actuators of the tailpipe must be controlled to ensure normal operation during varying conditions. Throughout the entire flight envelope, the tailpipe faces exceptionally harsh operating environments. Performance degradation is inevitable with increasing operating cycles, and the actuators inevitably undergo various forms of degradation under high temperature and high pressure. When degradation parameters change only slightly, simple control methods are insufficient to maintain the original dynamic performance when the controlled object changes. Therefore, an adaptive control system based on degradation conditions is required to ensure the safety and reliability of the tailpipe actuators. When the hydraulic components of the tailpipe actuators age, the integral time constant of the bearing displacement increases, resulting in gain degradation and a longer system response time, preventing timely attainment of steady-state values. Because the temperature, level, and pressure of the actuators all exhibit pure time lag, delayed degradation occurs after a certain period of operation, leading to reduced system stability and decreased dynamic performance during transient processes. Therefore, it is necessary to perform corresponding fault detection on this fault, or design a more robust controller to improve the fault tolerance of the system, so that the engine performance can be utilized more efficiently, thereby improving the overall performance of the aircraft.

[0003] To address the issue of parameter uncertainty caused by load variations in the system during repeated cyclic operation, which leads to interference signals and degradation of the controlled object, the teams of Turso [Robust Control of Deteriorated Turbofan Engines via Linear Parameter Varying Quadratic Lyapunov Function Design [C]] and Yue Xin [Asymptotic Tracking Control of Electro-hydraulic Load Simulator Based on Integral Robustness [J]] designed an adaptive active fault-tolerant controller that combines parameter adaptation with integral robustness control, using a hydraulic actuator as the controlled object. This improved the position tracking capability of the control effect and increased the fault tolerance of the system. The team led by Shao Wenxin [Optimization of BP Fuzzy Neural Network for Aero-engine Control Using Fast Simulated Annealing Algorithm [J]] used BP fuzzy neural network to optimize and adjust the PID controller of aero-engine actuators. Ding Kaifeng [Adaptive Control of Aero-engines Based on Adaline Network [J]] used a two-layer linear Adaline network to achieve online control of the engine within the entire flight envelope, but it still has problems such as complex structure, long adjustment time, and the need for real-time updating of controller parameters. Liu Xiaoyu [Model-Free Adaptive Aero-engine Control and Verification [D]] used PI control combining proportional control and anti-saturation method to achieve adaptive control of aero-engine models with actuator degradation. Zhang Tianhong [Fault-Tolerant Control for Aero-engine Component Performance Degradation [J]] et al. proposed an active fault-tolerant control design for component performance degradation based on sliding mode controller, which makes the engine have good dynamic characteristics, but the control parameters need to be reconfigured in the application process. Ji Xiaodong [Research on Transient Control of Aero-engines Based on ADRC [D]] et al. used ADRC-based transient closed-loop control of aero-engines to improve the flexibility of aero-engine acceleration / deceleration process, but did not directly control the actuators.

[0004] Existing technologies all suffer from reduced control accuracy due to different types of degradation of the tail nozzle actuator. Therefore, in order to solve this problem, it is necessary to conduct research on control methods with adaptive capabilities. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the problem of reduced control accuracy caused by different types of degradation of the tail nozzle actuator in the prior art, and to provide an adaptive control method for the engine tail nozzle actuator, which can still maintain high control accuracy when the tail nozzle actuator undergoes various types of degradation such as gain degradation and delay degradation.

[0006] An adaptive control method for an engine exhaust nozzle actuator is disclosed. This method uses a second-order linear LADRC controller to control the actuator and employs an improved Smith predictor to tune the system output y of the actuator. Finally, the tuned system output y′ from the improved Smith predictor is fed back to the second-order linear LADRC controller as the input signal to the state observer ESO in the LADRC controller. The improved Smith predictor comprises a conventional Smith predictor and a first-order filter. The output y1 of the conventional Smith predictor's built-in system model after a time delay is compared with the output y of the controlled object. p The difference dy between the two values ​​is filtered by the first-order filter to compensate for model inaccuracies. The output y of the traditional Smith predictor's built-in model does not go through a time delay element. p The sum of the output signal of the first-order filter and the output signal of the first-order filter is used as the tuned system output y′ of the improved Smith predictor output.

[0007] Preferably, the pole placement method is used to tune the parameters of the second-order linear LADRC controller.

[0008] Based on the same inventive concept, the following technical solutions can also be obtained:

[0009] An adaptive control device for an engine exhaust nozzle actuator, comprising:

[0010] A second-order linear LADRC controller is used to control the engine exhaust nozzle actuator; an improved Smith predictor is used to tune the system output y of the engine exhaust nozzle actuator, and the tuned system output y′ of the improved Smith predictor is fed back to the second-order linear LADRC controller as the input signal of the state observer ESO in the LADRC controller; the improved Smith predictor includes a conventional Smith predictor and a first-order filter, and the output y1 of the system built-in model of the conventional Smith predictor after time delay is compared with the output y of the controlled object. p The difference dy between the two values ​​is filtered by the first-order filter to compensate for model inaccuracies. The output y of the traditional Smith predictor's built-in model does not go through a time delay element. p The sum of the output signal of the first-order filter and the output signal of the first-order filter is used as the tuned system output y′ of the improved Smith predictor output.

[0011] Preferably, the pole placement method is used to tune the parameters of the second-order linear LADRC controller.

[0012] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0013] This invention addresses the problem of reduced control accuracy caused by different types of degradation in the actuator of the engine exhaust nozzle. It improves the traditional Smith predictor and organically combines it with the LADRC controller, which can ensure that the system can still maintain high control accuracy when the actuator experiences various types of degradation such as gain degradation and delay degradation. Actual experiments have verified that the adaptive control method proposed in this invention can eliminate the influence of disturbances from the external system more quickly, has stronger stability than PID control, and can achieve good dynamic command tracking when the system experiences internal disturbances. Compared with PID control, it has stronger applicability and robustness. Attached Figure Description

[0014] Figure 1 Diagram of the control loop for the cross-sectional area of ​​the nozzle throat;

[0015] Figure 2 Control structure diagram for the improved Smith predictor;

[0016] Figure 3 This is a control structure diagram of the control device of the present invention;

[0017] Figure 4 The PID control effect before and after adding the improved Smith predictor is shown in the figure for non-degradation control.

[0018] Figure 5(a) shows the PID control effect before and after adding the improved Smith predictor when the 20ms delay degrades;

[0019] Figure 5(b) shows the PID control effect before and after adding the improved Smith predictor when there is a 40ms delay degradation.

[0020] Figure 6 The LADRC control effect before and after adding the improved Smith predictor in the case of no degradation is shown in the figure.

[0021] Figure 7(a) shows the LADRC control effect before and after adding the improved Smith predictor when the 20ms delay degrades;

[0022] Figure 7(b) shows the LADRC control effect before and after adding the improved Smith predictor when there is a 40ms delay degradation.

[0023] Figure 8 A comparison of the control effects of LADRC and PID with an improved Smith predictor added when there is no degradation.

[0024] Figure 9 A comparison of the control effects of LADRC and PID with an improved Smith predictor added for step disturbances;

[0025] Figure 10(a) shows a comparison of the control effects of LADRC and PID with an improved Smith predictor added when the gain degrades by 30ms.

[0026] Figure 10(b) shows a comparison of the control effects of LADRC and PID with an improved Smith predictor added when the gain degrades by 40ms. Detailed Implementation

[0027] To address the problem of reduced control accuracy caused by different types of degradation in the actuator of the engine exhaust nozzle, the present invention proposes to improve the traditional Smith predictor and organically combine it with the LADRC controller, so as to ensure that the system can still maintain high control accuracy when the actuator experiences various types of degradation such as gain degradation and delay degradation.

[0028] The specific technical solution proposed in this invention is as follows:

[0029] An adaptive control method for an engine exhaust nozzle actuator is disclosed. This method uses a second-order linear LADRC controller to control the actuator and employs an improved Smith predictor to tune the system output y of the actuator. Finally, the tuned system output y′ from the improved Smith predictor is fed back to the second-order linear LADRC controller as the input signal to the state observer ESO in the LADRC controller. The improved Smith predictor comprises a conventional Smith predictor and a first-order filter. The output y1 of the conventional Smith predictor's built-in system model after a time delay is compared with the output y of the controlled object. p The difference dy between the two values ​​is filtered by the first-order filter to compensate for model inaccuracies. The output y of the traditional Smith predictor's built-in model does not go through a time delay element. p The sum of the output signal of the first-order filter and the output signal of the first-order filter is used as the tuned system output y′ of the improved Smith predictor output.

[0030] To facilitate public understanding, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings:

[0031] like Figure 1 As shown, the existing engine exhaust nozzle throat cross-sectional area is controlled by multiple actuators, resulting in a relatively large overall cavity. If directly driven by an electro-hydraulic servo valve, a flow rate exceeding 100L / min would be required, leading to its large size and weight. Therefore, an electro-hydraulic servo valve typically controls a distributor valve, which in turn drives the nozzle actuator. To ensure stability margin, the distributor valve employs closed-loop position control. Ignoring higher-order and nonlinear effects, the actuator loop of the distributor valve and exhaust nozzle is simplified to an integral element. The control law for the actuator loop of the distributor valve and exhaust nozzle is an amplification element, yielding the transfer function of the exhaust nozzle control loop:

[0032]

[0033] Therefore, the nozzle circuit is a second-order system. Where, K... PD K is the amplification factor of D8 for the A8 loop controller. pn For L n Loop controller amplification factor, K ln K is the amplification factor of the electro-hydraulic servo valve. D K is the integral time constant of the nozzle actuator. DA K is the conversion factor from the actuating volume of the actuator cylinder to the cross-sectional area of ​​the nozzle throat. nV K is the feedback amplification factor of the oil distributor valve circuit. DV This refers to the feedback amplification factor of the nozzle actuator. As the actuator is used more frequently, it tends to take longer to reach the designated position under the same input signal, a phenomenon known as gain degradation. Therefore, gain degradation can be addressed by reducing the amplification factor K of the electro-hydraulic servo valve. ln This can be achieved by increasing T. From formula (1), it can be concluded that this can be achieved by increasing T. n The size can achieve gain degradation.

[0034] In actual industrial production, control processes often exhibit varying degrees of lag. Lag is the delay caused by the limited transmission speed during the transmission of materials, energy, or signals. Pure lag generally refers to lag caused by transmission speed limitations, and in Simulink simulations, it is typically represented by a delay element.

[0035] To address the issue of time lag, existing controllers typically employ a Smith predictor structure, using a compensator connected in parallel with the controlled object to weaken or eliminate the pure time lag. However, because the built-in model used by the Smith predictor does not match the controlled object, the signal tuned by the Smith predictor will have a significant error compared to the ideal signal. Therefore, traditional Smith predictor control has poor robustness.

[0036] Therefore, this invention improves upon the traditional Smith predictor, and the structural principle of the improved Smith predictor is as follows: Figure 2 As shown, P is the controlled object, representing the transfer function G of the tail nozzle actuator. p (s) and the existing signal transmission hysteresis element e -τs .like Figure 2 As shown, based on the traditional Smith predictor, the improved Smith predictor uses the output y1 of the system's built-in model after the time delay element and the output y of the controlled object. pThe difference dy between the two signals is filtered and processed to analyze the output signal of the actuator. Based on the control signal u output by the controller, the system output y without time delay is obtained using a built-in model that does not contain time delay elements. p And using a filter F that can compensate for model misalignment, the output y1 of the system's built-in model after the time delay element and the output y of the controlled object are compared. p The difference dy between the two values ​​is tuned, and the output signal of filter F is compared with y. p The sum of these two signals, together with the feedback signal y' from the improved Smith predictor output, is applied to the total error signal e.

[0037] With the addition of the improved Smith predictor, the transfer function of the control loop is:

[0038]

[0039] It can be seen that when the Smith predictor's built-in model perfectly matches the controlled object, the signal after passing through filter F can be neglected, and the hysteresis is moved outside the transfer function and is not within the closed-loop control loop. Therefore, the system's control performance can maintain excellent performance. However, when the signal transmission delay time becomes longer, or when the model degrades and the controlled object does not match the Smith predictor's built-in model, the signal is filtered as a low-frequency interference signal through the first-order filter F, reducing the Smith predictor's sensitivity to modeling errors.

[0040] Based on the controlled object model established above, it is known that the model is a second-order object with a single input and output. Therefore, the improved Smith predictor is combined with a second-order linear LADRC controller for its control. The LADRC controller is an existing technology, consisting of a nonlinear feedback PD controller and an extended observer (ESO). It does not depend on the controlled object model and uses the extended observer to react to internal and external disturbances of the object. It is a control method that generally uses the bandwidth method for parameter tuning.

[0041] The overall control structure proposed in this invention, which combines the improved Smith predictor with a second-order linear LADRC controller, is as follows: Figure 3 As shown. In Figure 3In the diagram, the setpoint r is the input signal of the controller, d is the disturbance signal, and the two inputs of ESO are the control signal u and the system output y. The output signal of the actuator is first tuned by the improved Smith predictor. As can be seen from formula (2), after tuning by the improved Smith predictor, the hysteresis element will be moved outside the closed loop and will not affect the control effect of the added controller. According to the displacement theorem of Laplace transform, the pure time delay characteristic has no effect on the output response of the model. The output signal of the model is shifted by the time delay, and its waveform and dynamic performance remain the same. Therefore, using the output signal y' of the actuator tuned by the improved Smith predictor as the input signal of ESO in the LADRC controller can alleviate the influence of the hysteresis element on the system and enable the LADRC controller to obtain better control effect. ESO is the core of the controller and can estimate the internal and external disturbances of the system in real time. Z1, Z2, and Z3 are the three outputs of ESO. p K d b are controller parameters.

[0042] From formula (1), it can be seen that the controlled object module To facilitate parameter tuning for the LADRC algorithm, the transfer function is rewritten as a general form of differential equation:

[0043]

[0044] Since the system will degrade, taking into account the model changes caused by degradation and external interference signals, equation (3) can be rewritten as:

[0045]

[0046] in, The combined characteristics of internal and external disturbances in the system are represented by f = g + (b0 - b)u, which represents the extended state of the system. The mathematical expressions for f and g are not required to be known during parameter tuning and control.

[0047] It can be concluded that Figure 3 The control law is:

[0048]

[0049] Then formula (4) can be rewritten as:

[0050]

[0051] As the core of the LADRC controller, the ability of the ESO to observe states and estimate disturbances directly determines the control performance, and the level of the ESO's ability to observe states is determined by β. 01 ,β 02 ,β 03The design decision. The state equation of the ESO system is obtained as follows:

[0052]

[0053] In the formula, x3 = f represents the expanded state of the system. For an unknown disturbance that is not modeled in the system, f can be estimated using a state-space equation, expressed as:

[0054]

[0055] The linearly extended state observer can be further represented as:

[0056]

[0057] In the formula, β 01 ,β 02 ,β 03 It is the observer vector, which can be obtained by configuring poles.

[0058] Combining equation (4), it can be seen that by introducing an extended state observer, f will quickly converge to Z3, and the output of the system before and after degradation will quickly converge to the output of a second-order integral system. When the ESO can be correctly tuned, z1, z2, and z3 will track y respectively. f. To enable the system to achieve the desired control performance, this invention preferably employs the pole placement method for tuning the parameters of the LADRC controller.

[0059] To verify the effectiveness of the improved Smith predictor in mitigating signal transmission hysteresis, the control effects of adding an LADRC controller and a PID controller before and after the improved Smith predictor were compared. The comparison results are as follows: Figures 4 to 7(b) As shown in the figure, Smith represents the improved Smith predictor.

[0060] like Figure 4 As shown, under non-degradation conditions, the tail nozzle actuator exhibits excellent tracking performance under PID control and combined control of PID and an improved Smith predictor. Both systems have a settling time of less than 0.4s, no steady-state error, and good dynamic performance.

[0061] When lag degradation occurs during simulation, due to the hysteresis of the actuator's actuation system, the parameters of the TransportDelay module at the actuator output end are set to 40ms and 60ms to simulate the lag degradation present in the actuator. As shown in Figures 5(a) and 5(b), as the lag degradation of the nozzle actuator worsens, under single PID control, the rise time gradually increases, the overshoot gradually increases, and the response curve gradually fluctuates. However, under the improved Smith predictor and PID combined control, the rise time remains at 0.345s with no overshoot.

[0062] like Figure 6 As shown, under non-degradation conditions, the tail nozzle actuator exhibits excellent system tracking performance under LADRC control and combined control of LADRC and the improved Smith predictor, with settling times within 0.4s, no steady-state error, and good system dynamic performance.

[0063] As shown in Figures 7(a) and 7(b), with the increase of signal transmission hysteresis, under single LADRC control with unchanged control parameters, the rising phase of the response curve gradually increases, making it difficult to reach a steady-state value. However, under the combined control of the improved Smith predictor and LADRC, the settling time remains within 0.4s, and the response curve is stable. This demonstrates that even with the addition of the improved Smith predictor, both controllers can still meet the control targets well despite delay degradation, proving that the improved Smith predictor can effectively compensate for the hysteresis occurring during signal transmission.

[0064] The experiments above demonstrate that incorporating the Smith predictor effectively compensates for the delay degradation of the nozzle actuator. Simulations show that the improved Smith predictor increases the adjustment range of the LADRC's adjustable parameters. Therefore, the control performance of a PID controller with the improved Smith predictor and an LADRC controller are compared under different degradation conditions.

[0065] Under normal operating conditions, the output results of the nozzle actuator under two control modes after incorporating the improved Smith predictor are as follows: Figure 8 As shown, the control effect curve of LADRC has a high degree of overlap with that of PID, with a rise time of 0.33s and no overshoot.

[0066] Figure 9 The output responses of the two controllers are compared under external disturbances. It can be seen that, combined with the improved Smith predictor, the LADRC controller can eliminate the influence of external disturbance signals more quickly than the PID controller, enabling the system to reach stability faster.

[0067] As the actuator is used more frequently, it tends to take longer to reach the designated position under the same input signal, a phenomenon known as gain degradation. Therefore, gain degradation can be addressed by reducing the amplification factor K of the electro-hydraulic servo valve. ln To achieve this, we can derive from formula (1) that increasing T... n Gain degradation was simulated for 30ms and 40ms. Figures 10(a) and 10(b) show a comparison of the output responses of the two controllers under gain degradation. It can be seen that as the degree of gain degradation increases, the settling time of the PID controller combined with the improved Smith predictor gradually increases to 0.48s and 0.76s respectively, resulting in poorer command tracking performance. However, the LADRC controller combined with the improved Smith predictor maintains its original dynamic performance even with increased degradation, indicating that the LADRC control method has stronger applicability and robustness in adaptive control of nozzle actuators.

Claims

1. An adaptive control method for an engine exhaust nozzle actuator, characterized in that, A second-order linear LADRC controller is used to control the engine exhaust nozzle actuator, and an improved Smith predictor is used to tune the system output y of the actuator. Finally, the tuned system output y′ from the improved Smith predictor is fed back to the second-order linear LADRC controller as the input signal to the state observer ESO in the LADRC controller. The improved Smith predictor includes a conventional Smith predictor and a first-order filter. Based on the control signal u output by the controller, the system output y without time delay is obtained using the built-in system model of the conventional Smith predictor, which does not contain a time delay element. p The output y1 is obtained after passing through a time delay element; the output y1 of the system built-in model of the traditional Smith predictor without a time delay element is compared with the output y1 of the controlled object after passing through the time delay element. p The difference dy between the two values ​​is filtered by the first-order filter to compensate for model inaccuracies. The traditional Smith predictor's built-in model without a time delay element outputs y without a time delay element. p The sum of the output signal of the first-order filter and the tuned system output y′ of the improved Smith predictor output is applied to the total error signal e.

2. The adaptive control method for the engine exhaust nozzle actuator as described in claim 1, characterized in that, The pole placement method is used to tune the parameters of the second-order linear LADRC controller.

3. An adaptive control device for an engine exhaust nozzle actuator, characterized in that, include: A second-order linear LADRC controller is used to control the engine exhaust nozzle actuator; An improved Smith predictor is used to tune the system output y of the engine exhaust nozzle actuator, and the tuned system output y′ output by the improved Smith predictor is fed back to the second-order linear LADRC controller as the input signal of the state observer ESO in the LADRC controller. The improved Smith predictor includes a conventional Smith predictor and a first-order filter. Based on the control signal u output by the controller, the system output y without time delay is obtained using the built-in system model of the conventional Smith predictor without time delay. p The output y1 is obtained after passing through a time delay element; the output y1 of the system built-in model of the traditional Smith predictor without a time delay element is compared with the output y1 of the controlled object after passing through the time delay element. p The difference dy between the two values ​​is filtered by the first-order filter to compensate for model inaccuracies. The traditional Smith predictor's built-in model without a time delay element outputs y without a time delay element. p The sum of the output signal of the first-order filter and the tuned system output y′ of the improved Smith predictor output is applied to the total error signal e.

4. The adaptive control device for the engine exhaust nozzle actuator as described in claim 3, characterized in that, The pole placement method is used to tune the parameters of the second-order linear LADRC controller.

Images