An aviation flap PMSM multi-source heterogeneous fusion position information reconstruction method and system

Abstract

The application discloses a kind of aviation flap PMSM multi-source heterogeneous fusion position information reconstruction method and system, relating to the control technical field of aviation flap electromechanical actuator, the kinematics model of flap, improved high-frequency square wave injection method, extended Kalman filter and extended H ∞ filter double observer heterogeneous fusion are combined, and three-source heterogeneous redundant reconstruction architecture is constructed, and through full-speed domain weighted fusion strategy, complementary advantages are realized, to provide high-precision, high-reliability motor rotation angle and speed state reconstruction for flap three closed-loop control, and then flap angle is calculated by kinematics model. It is suitable for full-speed domain operating scene to realize high-precision, high-robustness state observation, and provides accurate control information for the closed-loop control of flap deflection angle.

Description

Technical Field

[0001] This invention relates to the field of control technology for electro-mechanical actuators (EMA) of aircraft flaps, and more specifically to a method and system for reconstructing position information from multiple heterogeneous sources using a permanent magnet synchronous motor (PMSM) for aircraft flaps. Background Technology

[0002] The aircraft flap (EMA) is a critical actuator in the aircraft's high-lift system, and its control accuracy and reliability directly affect flight safety. In modern aircraft EMAs, permanent magnet synchronous motors widely employ a rotor field-oriented vector control strategy. The core prerequisite for this strategy is the real-time acquisition of high-precision rotor position and speed information to achieve current decoupling and torque control. Without accurate position information, vector control cannot be established, leading to anything from current runaway and torque pulsation to flap positioning failure, threatening flight safety. Therefore, under any operating condition, even when sensors may fail or signals are weak, continuously obtaining high-precision rotor position information is a prerequisite for achieving high-performance flap control.

[0003] Currently, the mainstream methods for obtaining rotor position include direct measurement using physical sensors and position observation based on back electromotive force (EMF). Physical sensors offer high accuracy under normal operating conditions, but in environments with strong airborne vibrations, wide temperature ranges, and strong electromagnetic interference, the sensors themselves are high-failure-rate components. If a wire breaks, jams, or signal drifts, the system will directly lose position feedback. The back EMF-based observation method performs well under medium-to-high-speed conditions, but it has a fundamental physical limitation: the back EMF amplitude is directly proportional to the motor speed. Flange systems operate extensively in low-speed fine-tuning positioning and zero-speed holding conditions, where the back EMF signal is extremely weak, causing a sharp deterioration in the signal-to-noise ratio and leading to observation failure. Therefore, in situations where sensor signals are scarce or the back EMF is weak, a state reconstruction algorithm independent of these conditions must be introduced as a redundancy or alternative.

[0004] In the field of permanent magnet synchronous motor state estimation, researchers have proposed various rotor position observation methods. However, these methods each have inherent limitations when applied to aircraft flap electromechanical actuators. The first category is methods based on back electromotive force (EMF) estimation, such as sliding mode observers and model reference adaptive methods. Their core principle is to use the back EMF generated during motor rotation to estimate the rotor position. These methods perform well at medium and high speeds, but at low and zero speeds, the back EMF amplitude decreases sharply with decreasing speed, severely degrading the signal-to-noise ratio and leading to a significant drop in position estimation accuracy or even failure. The second category is the traditional high-frequency square wave injection method. Although it can operate at zero and extremely low speeds, it suffers from three main problems: reliance on filters to separate high-frequency current leads to phase delay and amplitude attenuation; traditional phase-locked loops or Romberg observers experience increased errors under load disturbances; and signal separation is difficult and leads to failure at medium and high speeds. The third category is the traditional Romberg observer. Although simple in structure and computationally inexpensive, it typically assumes a known constant load torque, resulting in a significant increase in reconstruction error when the load changes, and it is also highly sensitive to changes in motor parameters. A single observer cannot simultaneously meet the comprehensive requirements of full-speed range, high dynamics, and strong robustness in aviation. In addition, existing methods often separate motor state reconstruction from flap kinematic models, failing to form an integrated electromechanical coupling reconstruction framework.

[0005] Therefore, in view of the shortcomings of the existing technology, how to provide a method and system for reconstructing the multi-source heterogeneous fusion position information of aircraft flap PMSM is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a method and system for reconstructing position information of aero-aircraft flap PMSM through multi-source heterogeneous fusion. It combines a flap kinematic model, an improved high-frequency square wave injection method, and a heterogeneous fusion of extended Kalman filtering and extended H∞ filtering dual observers to construct a three-source heterogeneous redundant reconstruction architecture. A full-speed-domain weighted fusion strategy is used to achieve complementary advantages, providing high-precision and high-reliability motor angle and speed state reconstruction for the three-loop control of the flap. The flap angle is then calculated using the kinematic model. This method is suitable for achieving high-precision and high-robustness state observation in full-speed-domain operating scenarios, providing accurate control information for closed-loop control of the flap deflection angle.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a method for reconstructing the position information of an aircraft flap PMSM through multi-source heterogeneous fusion, comprising: Collect multi-source aviation signals and preprocess the aviation signals to generate standardized data; A heterogeneous redundancy architecture combining physical and analytical methods is used to perform parallel and independent computation on the standardized data to obtain multi-source redundant signals; Different operating conditions are divided according to the complete flight mission profile of the flaps. The multi-source redundant signals are subjected to condition-adaptive weighted fusion to output the reconstructed values ​​of the motor position and speed. The pre-established kinematic model of the flap electromechanical actuator is called to solve the reconstructed position and speed values ​​to obtain the actual deflection angle of the flap, and the final motor rotation angle, speed and flap angle are output.

[0008] Preferably, a heterogeneous redundancy architecture combining physical and analytical methods is used to perform parallel and independent computation on the standardized data, including: Source A is the measured rotor position signal directly acquired by a physical sensor; Source B employs dual observers, EKF and EH∞, and outputs medium-to-high-speed analytical position observation signals through confidence assessment and heterogeneous fusion. Source C employs an improved high-frequency square wave injection method, extracts position errors through filterless current separation and vector cross multiplication, and combines an improved Romberg observer that incorporates load torque estimation to output a zero-low-speed analytical position observation signal.

[0009] Preferably, the operating conditions include a zero-low speed range, a medium-high speed range, and a transition speed range; The zero-low speed domain is determined based on the observation results of source C, the medium-high speed domain is determined based on the observation results of source B, and the transition speed domain is weighted and fused by nonlinear weights optimized offline by an improved differential evolution algorithm to output the final position and speed reconstruction values.

[0010] Preferably, source C employs an improved high-frequency square wave injection method, extracts position errors through filterless current separation and vector cross product, and combines an improved Romberg observer incorporating load torque estimation to reconstruct the output position under flap low-speed fine-tuning positioning and zero-speed holding conditions, including: A high-frequency model of a built-in permanent magnet synchronous motor under high-frequency excitation is established by alternately injecting constant-amplitude high-frequency square wave excitation voltage into the rotating coordinate system of the motor. By utilizing the fundamental current characteristics at three adjacent discrete sampling times, the high-frequency response current along the α-β axis is obtained by solving the current equations at the three sampling points simultaneously. Based on the high-frequency model, under square wave injection, the current change in each sampling period is calculated to obtain the high-frequency incremental current; The error signal is extracted from the separated high-frequency incremental current using the vector cross product method; Rotor reconfiguration position, reconfiguration speed, and load torque are selected as state variables. A Romberg observer based on load disturbance compensation is constructed, and the observation state is corrected according to the error signal to output the rotor position and speed observation values ​​under zero low-speed conditions.

[0011] Preferably, a high-frequency model of the built-in permanent magnet synchronous motor under high-frequency excitation is established by alternately injecting constant-amplitude high-frequency square wave excitation voltages into the rotating coordinate system of the motor estimation, including: The voltage equation of the built-in permanent magnet synchronous motor in the dq rotating coordinate system is: ; Where p = d / dt, R s L is the stator resistance. d , L q These are the inductances along the d and q axes, respectively, where ω is the electric angular velocity and ψ is f For permanent magnet flux linkage; The high-frequency voltage equation is expressed as: ; Give the estimated rotating coordinate system Injecting high-frequency square wave voltage: ; Among them, V h Let (-1) represent the magnitude of the injected voltage. k The alternation of symbols corresponds to the injection frequency. f h = f PWM / 2, sampling period T s =1 / f PWM ; By transforming the coordinates, we obtain the high-frequency model in the stationary a-β coordinate system: ; in, ;θ e This is the actual rotor electrical angle. The reconstructed rotor electrical angle, reconstruction error .

[0012] Preferably, a dual observer architecture using EKF and EH∞ is employed, achieving a complementary balance between normal high dynamics and anomaly robustness through confidence assessment and heterogeneous fusion, including: A PMSM model is constructed in the α-β stationary coordinate system. The process noise and measurement noise are assumed to follow a Gaussian distribution. Prediction-update iterations are completed through Euler discretization and Jacobian matrix solving. A chi-square consistency test is constructed based on the observed information sequence, and the EKF confidence level γ is quantitatively calculated. EKF Output the rotor position observation results under normal flight conditions; The back EMF, inductance, and resistance are simultaneously reconstructed as state variables to build an augmented state model. The H∞ filtering cost function is iteratively solved, and the rotor position is extracted from the augmented state. The EH∞ confidence level γ is calculated using root mean square quantization of the residuals through a sliding window. EH∞ Output the rotor position observation results under motor parameter perturbation and electromagnetic interference conditions; Based on EKF confidence level γ EKF and EH∞ confidence level γ EH∞ Four working modes are defined: EKF normal dominant mode, EH∞ arbitration mode, transition zone weighted fusion mode, and dual observer untrusted mode. The observer fault status flag is output synchronously, and the single-channel medium and high speed rotor position observation value is fused and output.

[0013] Preferably, the EKF confidence level γ EKF The calculation process is as follows: Based on chi-square statistics To define the EKF confidence level: ; EH∞ confidence level γ EH∞ The calculation process is as follows: Based on the sliding window residual RMS, the confidence level of EH∞ is: ; Where, δ H∞,max The maximum allowed residual RMS threshold.

[0014] Preferably, the improved differential evolution algorithm employs parameter adaptation, dynamically adjusting the mutation factor F and crossover probability CR in each generation based on the success rate, as follows: NP individuals are uniformly generated in the λ1 space, where λ1∈[0,1] and λ2=1-λ1; ; For the t-th generation individual Generate mutated individuals: ; Where r1, r2, and r3 are distinct random indices, and the scaling factor F is adaptively adjusted: ; Inherited components from mutated individuals using crossover probability CR: ; Crossover probability CR adaptive adjustment: ; Based on the fitness function, the best candidate is selected: ; The fitness function minimizes the sum of squared weighted speed prediction errors: ; Where M is the number of different speed points within the switching speed domain, ω true,j The actual rotational speed (obtained through offline calibration of a high-precision encoder); Update the next generation : ; ; The optimal weighting coefficients for different speeds were obtained through offline optimization.

[0015] Preferably, an aircraft flap PMSM multi-source heterogeneous fusion position information reconstruction system includes: The physical layer is used to acquire multi-source aviation signals and preprocess the aviation signals to generate standardized data. The analytical redundancy layer is used to perform parallel and independent computation on the standardized data using a heterogeneous redundancy architecture that combines physical and analytical methods to obtain multi-source redundant signals. The full-speed domain fusion decision layer is used to divide different operating conditions according to the complete flight mission profile of the flaps, perform condition-adaptive weighted fusion of the multi-source redundant signals, and output the reconstructed values ​​of the motor position and speed. The output layer is used to call the pre-established kinematic model of the flap electromechanical actuator, solve the reconstructed position and speed values ​​to obtain the actual deflection angle of the flap, and output the final motor rotation angle, speed and flap angle.

[0016] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method and system for reconstructing the position information of a PMSM multi-source heterogeneous fusion of aerospace flaps. In view of the high precision and high reliability requirements of the rotor position state reconstruction of the permanent magnet synchronous motor under the full speed domain operation conditions of the aerospace flap EMA, the present invention proposes a full speed domain state reconstruction method that combines an improved high-frequency square wave injection method with heterogeneous fusion of dual observers, providing accurate motor state feedback for the three-closed-loop control of the flap angle.

[0017] First, a kinematic model of the flap electromechanical actuator is established. Based on physical parameters such as reduction ratio, roller screw lead, and connecting rod lever arm, the transmission relationship from motor rotation angle to flap deflection angle is derived, providing physical constraints and engineering solution basis for state reconstruction.

[0018] Secondly, a multi-source heterogeneous redundant position state reconstruction architecture is constructed. Three independent sources of position information are defined: direct measurements from physical sensors θsen (source A), and heterogeneous fusion outputs from the EKF and EH∞ dual observers. (Source B) Improved high-frequency square wave injection sensorless observation values (Source C). The three position sources are calculated independently in parallel and cross-checked to provide highly reliable motor position information support for flap deflection control.

[0019] Then, a high-precision rotor position observation in the zero-low-speed domain is achieved by improving the high-frequency square wave injection method. This method achieves filterless current separation based on algebraic operations on three adjacent sampling points, and introduces an improved Romberg observer with load torque as an extended state, obtaining rotor position observations independent of physical sensors in the zero-low-speed domain. As an independent source of parsing redundancy in multi-source information.

[0020] Simultaneously, a heterogeneous fusion architecture with dual observers is constructed. Using an extended Kalman filter (EKF) as the master observer, it provides high-bandwidth, low-latency state reconstruction under normal conditions. Using an extended H∞ filter (EH∞) as the arbitrator observer provides a more robust reconstruction when model parameter perturbations or noise statistics are unknown. The two are output through confidence assessment and weighted fusion. This achieves the complementary advantages of "high dynamism in normal conditions and strong robustness in abnormal conditions".

[0021] Finally, a multi-source weighted fusion strategy is designed for the entire speed domain. The multi-source positions are weighted and fused according to the speed range: in the zero-low speed domain, improved high-frequency injection (source C) is the main method; in the medium-high speed domain, dual-observer fusion (source B) is the main method; and in the transition speed domain, a smooth transition is achieved through nonlinear weights optimized by an improved differential evolution algorithm. The final output is the reconstructed motor rotation angle value. As position feedback for closed-loop control of flap deflection angle, and by solving the reconstructed value of flap deflection angle through a kinematic model. .

[0022] This method provides high-precision and high-reliability drive motor state reconstruction for the three-closed-loop control of aerospace flap electromechanical actuators through multi-dimensional innovations such as flap kinematic modeling, multi-source heterogeneous redundancy architecture, improved high-frequency injection, complementary fusion of dual observers, and weighted fusion in the full-speed domain. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of a method for reconstructing the location information of an aircraft flap PMSM multi-source heterogeneous fusion according to an embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the flap actuator system architecture provided in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the flap connection provided in an embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of heterogeneous fusion of dual observers provided in an embodiment of the present invention.

[0028] Figure 5 This is a schematic diagram of full-speed domain fusion switching provided in an embodiment of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] This invention discloses a method for reconstructing the position information of an aircraft flap PMSM through multi-source heterogeneous fusion, comprising: Collect multi-source aviation signals and preprocess the aviation signals to generate standardized data; A heterogeneous redundancy architecture combining physical and analytical methods is used to perform parallel and independent computation on the standardized data to obtain multi-source redundant signals; Different operating conditions are divided according to the complete flight mission profile of the flaps. The multi-source redundant signals are subjected to condition-adaptive weighted fusion to output the reconstructed values ​​of the motor position and speed. The pre-established kinematic model of the flap electromechanical actuator is called to solve the reconstructed position and speed values ​​to obtain the actual deflection angle of the flap, and the final motor rotation angle, speed and flap angle are output.

[0031] The motor rotation angle, speed, and flap angle are fed into a three-loop control system to meet the flap angle requirements during takeoff, cruise, and landing.

[0032] The state reconstruction system proposed in this invention is based on the kinematic model of the flap electromechanical actuator and adopts a four-layer architecture consisting of five cascaded core modules. The overall data flow follows the logical sequence of "signal acquisition → multi-source parallel computing and full-speed domain fusion → state output", providing motor angle and speed feedback for flap closed-loop control.

[0033] Specifically, a heterogeneous redundancy architecture combining physical and analytical methods is used to perform parallel and independent computations on the standardized data, including: Source A is the measured rotor position signal directly acquired by a physical sensor; Source B employs a dual observer system of EKF and EH∞, and outputs medium-to-high-speed analytical position observation signals through confidence assessment and heterogeneous fusion, achieving a complementarity between normal high dynamics and anomaly robustness; Source C employs an improved high-frequency square wave injection method, extracts position errors through filterless current separation and vector cross multiplication, and combines an improved Romberg observer that incorporates load torque estimation to output a zero-low-speed analytical position observation signal, thereby outputting high-precision position reconstruction under flap low-speed fine-tuning positioning and zero-speed holding conditions.

[0034] The three sources are independent of each other, forming a heterogeneous redundant architecture that combines physical and analytical methods.

[0035] Specifically, it involves collecting multi-source aviation signals, including: Aviation signals are acquired using a rotary transformer, current and voltage sensors, and a PWM drive protection circuit that meets aviation power supply characteristics.

[0036] like Figure 2 As shown, the core components of the flap actuator mainly include a servo motor and its controller, a reducer and a ball screw. The motor output angle is converted into a linear displacement at the actuator output end through the reduction ratio (N) and the ball screw lead (P), thereby driving the flap to deflect.

[0037] Let the angular displacement and angular velocity of the motor rotor be θ. m and ω m According to the principle of gear transmission, the angular velocity ω at the output end of the reducer, i.e., the lead screw, is... s for: (1) There is an inherent kinematic coupling between the input and output of a roller screw pair. Let the lead of the screw be P, meaning that for every revolution of the screw, the nut advances one lead distance. Therefore, the relationship between the linear velocity v of the nut and the angular velocity of the screw is: (2) Integrating equation (2) and setting the initial position to zero, we can obtain the core transformation equation between the linear displacement x of the output rod and the motor rotation angle: (3) like Figure 3 As shown, the flap is connected to the trailing edge of the wing via a hinge, and the linear output end of the EMA is connected to the flap via a rigid connecting rod. Let the instantaneous lever arm length from the flap pivot O to the actuator connection point A be R. When the linear displacement is x, the connection point moves along the arc. If the change of R during the transmission process is relatively small, it can be approximated as a constant within the range of flap deflection angle changes.

[0038] Based on the circular motion relationship, the flap deflection angle θ caused by the linear displacement x of the actuator... f satisfy: (4) Combining equations (3) and (4), we obtain the value from the motor rotation angle θ. m To flap angle θ f transitive relationships: (5) Summarized as follows: (6) In engineering applications, angles are often measured in degrees (°). and These are the degree values ​​of the motor rotation angle and the flap angle, respectively, and they satisfy the following relationship with the radian values: (7) Substituting equation (6) and simplifying, we obtain the kinematic transmission model of the flap electromechanical actuation system: (8) From equation (8), the corresponding motor rotation angle can be obtained as: (9) The above relationship shows that by measuring the motor rotation angle θ m This allows us to indirectly obtain the flap movement angle θ. f This allows for precise adjustment of the flap position.

[0039] Specifically, the operating conditions include the zero-low speed domain, the medium-high speed domain, and the transition speed domain; The zero-low speed domain is determined based on the observation results of source C, the medium-high speed domain is determined based on the observation results of source B, and the transition speed domain is weighted and fused by nonlinear weights optimized offline by an improved differential evolution algorithm to output the final position and speed reconstruction values.

[0040] Ensure the continuity and reliability of flap position reconfiguration within the entire flight mission profile.

[0041] The flap electromechanical actuator includes: a servo motor and its controller, a reducer and a ball screw; the motor output angle is converted into a linear displacement at the actuator output end through the reduction ratio (N) and the ball screw lead (P), thereby controlling the flap deflection.

[0042] Specifically, source C employs an improved high-frequency square wave injection method, extracting position errors through filterless current separation and vector cross product, and combining this with an improved Romberg observer incorporating load torque estimation. This enables high-precision position reconstruction under low-speed flap fine-tuning and zero-speed holding conditions, including: Equal-amplitude high-frequency square wave excitation voltages are alternately injected into the rotating coordinate system of the motor. The frequency of the high-frequency voltage is much higher than the fundamental frequency of the motor. The stator resistance voltage drop and the rotating back electromotive force term are ignored. A high-frequency model of the built-in permanent magnet synchronous motor under high-frequency excitation is established. By taking advantage of the fact that the fundamental current is approximately constant and the high-frequency injection current alternates between positive and negative at three adjacent discrete sampling times, the high-frequency response current along the α-β axis can be obtained by solving the current equations at the three sampling points simultaneously. Based on the high-frequency model, under square wave injection, the current change in each sampling period is calculated to obtain the high-frequency incremental current; The error signal is extracted from the separated high-frequency incremental current using the vector cross product method; Rotor reconfiguration position, reconfiguration speed, and load torque are selected as state variables. A Romberg observer based on load disturbance compensation is constructed, and the observation state is corrected according to the error signal to output the rotor position and speed observation values ​​under zero low-speed conditions.

[0043] Specifically, the main function of the improved high-frequency square wave injection method is to provide analytical redundant position information independent of physical sensors in the zero-low velocity domain. Its core components include: high-frequency square wave injection, filterless current separation, vector cross-multiplication position error extraction, and an improved Romberg observer.

[0044] Specifically, a high-frequency model of a built-in permanent magnet synchronous motor under high-frequency excitation is established by alternately injecting constant-amplitude high-frequency square wave excitation voltages into the rotating coordinate system of the motor, including: In the mathematical model of PMSM under high-frequency excitation: The voltage equation of the built-in permanent magnet synchronous motor in the dq rotating coordinate system is: (10) Where p = d / dt, R s L is the stator resistance. d , L q These are the inductances along the d and q axes, respectively, where ω is the electric angular velocity and ψ is f For permanent magnet flux linkage; When the injected square wave voltage frequency ω h When the voltage drop across the resistor is much higher than the fundamental frequency ω (typically ωh / ω>10), the voltage drop across the resistor can be ignored. s i, the rotating electromotive force term ωLi, and ω, the high-frequency voltage equation can be simplified to: (11) High-frequency square wave injection: Give the estimated rotating coordinate system Injecting high-frequency square wave voltage: (12) Among them, V hThe magnitude of the injected voltage (typically 5% to 10% of the rated voltage), (-1) k The alternation of symbols corresponds to the injection frequency. f h = f PWM / 2, sampling period T s =1 / f PWM ; By transforming the coordinates, we obtain the high-frequency model in the stationary a-β coordinate system: (13) in, (14) θ e This is the actual rotor electrical angle. Define the reconfiguration error for the reconfigured rotor electrical angle. .

[0045] Specifically, the filterless current separation process includes: Traditional methods use bandpass filters to extract high-frequency currents, which introduces phase delay.

[0046] This invention employs a filterless separation method, where the α-axis response current under discrete sampling can be expressed as the sum of the high-frequency response current and the fundamental current: (15) Since the fundamental frequency is much lower than the square wave frequency, it can be assumed that the fundamental current remains constant within three adjacent sampling points: (16) Meanwhile, the high-frequency response current is the same in magnitude but opposite in direction at each sampling point (square wave injection characteristic): (17) Combining equations (15) and (17), we get: (18) Solving the system of equations yields the high-frequency response current: (19) Similarly, the β-axis high-frequency response current can be obtained: (20) This method requires no filters, eliminating phase delay and amplitude attenuation caused by filters. The method is also applicable to the dq rotating coordinate system.

[0047] Specifically, in this embodiment of the invention, high-frequency incremental current is used to calculate and extract the position error. The relationship between the derivative of the high-frequency current and the voltage can be obtained from equation (13). Under square wave injection, the current change in each sampling period, i.e., the high-frequency incremental current, is: (twenty one) The components of the injected voltage in the a-β coordinate system are: (twenty two) When PMSM rotor position reconstruction error When it converges to 0, Substituting equation (13) into equation (12) and simplifying, we get: (twenty three) in, .

[0048] Using the vector cross product method to extract the position error signal can avoid the sensitivity to noise caused by the arctangent operation.

[0049] Define the demodulation signal: (twenty four) Substituting equation (23) into equation (24), consider... For the general case where the result is not zero, retain the first-order approximation: (25) Multiplying by (-1)k and passing through a low-pass filter (i.e., with a cutoff frequency much lower than fh), we obtain the average error signal: (26) Among them, gain .

[0050] Specifically, this embodiment of the invention employs an improved Romberg observer, as traditional Romberg observers or PLLs experience increased errors under load disturbances. This embodiment of the invention introduces external load torque disturbances as extended state variables, thereby improving the observer structure.

[0051] The mechanical motion equation of PMSM is: (27) Where, ω m P is the mechanical angular velocity. n Let J be the pole logarithm, J be the moment of inertia, and T be the moment of inertia. e For electromagnetic torque, T L Where is the load torque, and B is the viscous friction coefficient.

[0052] Ignoring the derivative of the load torque With viscous friction B≈0, the rotor reconfiguration position is selected. Reconstructed speed and load torque For the state variables, construct the improved state equation: (28) in, The electromagnetic torque is given. K1, K2, and K3 are the position error signals and the observer gains.

[0053] Configure the observer poles according to the bandwidth ρ so that the observer response speed is about 1 / 5 to 1 / 10 of the current loop bandwidth.

[0054] (28)

[0055] This observer treats disturbances as independent states for reconstruction, effectively compensating for reconstruction errors caused by sudden load changes; the output is... and .

[0056] Specifically, this embodiment of the invention employs heterogeneous fusion of dual observers, consisting of EKF and EH∞ dual observers, and outputs the results through confidence evaluation and heterogeneous fusion. .

[0057] Specifically, using a dual observer architecture of EKF and EH∞, and achieving complementarity between normal high dynamics and anomaly robustness through confidence assessment and heterogeneous fusion, including: A PMSM model is constructed in the α-β stationary coordinate system. The process noise and measurement noise are assumed to follow a Gaussian distribution. Prediction-update iterations are completed through Euler discretization and Jacobian matrix solving. A chi-square consistency test is constructed based on the observed information sequence, and the EKF confidence level γ is quantitatively calculated. EKF It outputs high-bandwidth, low-latency rotor position observation results under normal flight conditions; The back EMF, inductance, and resistance are simultaneously reconstructed as state variables to build an augmented state model. The H∞ filtering cost function is iteratively solved, and the rotor position is extracted from the augmented state. The EH∞ confidence level γ is calculated using root mean square quantization of the residuals through a sliding window. EH∞ Output the rotor position observation results under motor parameter perturbation and electromagnetic interference conditions; Based on EKF confidence level γ EKF and EH∞ confidence level γ EH∞ Four working modes are defined: EKF normal dominant mode, EH∞ arbitration mode, transition zone weighted fusion mode, and dual observer untrusted mode. The observer fault status flag is output synchronously, and the single-channel medium and high speed rotor position observation value is fused and output.

[0058] Specifically, the embodiments of the present invention employ an extended Kalman filter (EKF) design, including: (1) State-space model The PMSM model is constructed in a stationary a-β coordinate system. The state variables are... The input is The output is .

[0059] Continuous-time state equations: (30) Among them, L s ≈L d ≈L q (Applicable to surface-mount motors or those after equivalent treatment; average inductance can be used for built-in motors), P n Let J be the pole pair number, J be the moment of inertia, B be the coefficient of viscous friction, and T be the coefficient of polarity. L The load torque is ω1~ω4, and the process noise is ω1~ω4, which typically follows a Gaussian distribution N(0,Q). EKF ).

[0060] The measurement equation is: (31) Where v1~v2 are measurement noise, which typically follows a Gaussian distribution N(0,R). EKF ).

[0061] (2) Discretization and Jacobian matrix

[0062] First-order Euler discretization is used, with a sampling time Δt=T. s : (32) (33) in: ; Component expressions for the nonlinear function fEKF: (34) (35) (36) (37) Find the Jacobian matrix for xk-1 : (38) in: (39) (40) (41) (3) EKF iteration initialization: (42) (43) Prediction phase: (44) (45) Calculate the new information: (46) (47) Update phase: (48) (49) (50) Output: (51) (4) Consistency test between the new information sequence and the chi-square test Define the innovation sequence: (52) Assuming the EKF is functioning correctly and the noise is Gaussian, the innovation sequence follows a zero-mean Gaussian distribution with a covariance of S. k : (53) Constructing the chi-square detection quantity: (54) in, It follows a chi-square distribution with 2 degrees of freedom, and two thresholds are set: Corresponding to the 90th percentile; Corresponding to the 99.9th percentile.

[0063] Specifically, the embodiments of the present invention employ an extended H∞ filter (EH∞) design, including: (1) Augmented state space model This invention employs an augmented state model, simultaneously reconstructing back electromotive force (EMF), inductance, and resistance as state variables, maintaining robustness even when noise statistical characteristics are unknown. The model is established in a stationary a-β coordinate system. Since the back EMF changes slowly, its derivative d is assumed. eα / d t =d eβ / d t=0, and also treat the inductor L and resistor R as state variables, resulting in the augmented state vector: (55) Input vector Measure the output vector .

[0064] Continuous-time state equations: (56) First-order Euler discretization is used, with a sampling time T. s The discrete state equations are obtained as follows: (57) (58) Where, ω k v k For process noise and measurement noise, the covariance matrices are Q and Q, respectively. k and R k However, precise statistical properties are not required. Measurement matrix: ; nonlinear functions f EH∞ ( x k-1 , u k-1 The explicit expression for ) is: (59) (2) Jacobian matrix calculation f ( x , u Jacobian matrix of x for: (60) All elements are in The value is taken at that location.

[0065] (3) Extended H∞ filter iteration

[0066] Define the cost function: (61) in: S represents the performance boundary (usually a small positive number, such as 10⁻⁶). k It is a positive definite symmetric weighted matrix.

[0067] The recursive formula for the extended H∞ filter is: initialization: (62) (63) Where L0 is the nominal value of inductance and R0 is the nominal value of resistance.

[0068] Prediction phase: (64) (65) Gain calculation: (66) Update phase: (67) (68) (4) Rotor position extraction Extracting the back electromotive force component from the augmented state: (69) The rotor position can be calculated using the arctangent, or a smoother position output can be obtained by extracting it using a phase-locked loop (PLL). (70) Speed ​​Reconstruction: (71) (5) Residual calculation and confidence index Define the reconstruction residual of EH∞: (72) Calculate the root mean square (RMS) of the residuals within the sliding window: (73) Where N is the window length (e.g., 100 sampling points, corresponding to 10ms).

[0069] (6) Proof of the boundedness of reconstruction error of extended H∞ filter

[0070] Theorem: For nonlinear systems (57)-(58) and extended H∞ filtering algorithms (64)-(68), if there exist positive real numbers that satisfy the following assumptions, then the state reconstruction error is... The mean square is bounded.

[0071] Assumption 1: There exist real numbers f min , f max , α min , α max , h min , hmax satisfy: (74) Assumption 2: There exist real numbers r min , r max , q min , q max , s min , s max satisfy: (75) Assumption 3: Positive real numbers exist. ε max >0 satisfies the bounded condition for linearization error.

[0072] prove: Define the Lyapunov function: (76) Depend on The positive definiteness of indicates that there exist positive real numbers. ν min , ν max Make: (77) The state error is dynamically substituted, expanded using Taylor, and then introduced into a diagonal matrix. α k Compensating for linearization error yields: (78) (79) Substituting into the update equation (77), the conditional expectation can be derived as follows: (80) in, Bounded. According to the boundedness theorem for stochastic processes, the reconstruction error... The mean square is bounded.

[0073] Specifically, dual-observer confidence assessment and heterogeneous fusion: EKF confidence level γ EKF The calculation process is as follows: Based on chi-square statistics To define the EKF confidence level: (81) in, =4.605, corresponding to 2 degrees of freedom, 90% confidence level; =13.815, corresponding to 2 degrees of freedom, 99.9% confidence level; When γ EKF When =1, the information is normal, indicating that EKF is working properly; when At times, if the innovation is too large, the EKF may be affected by model mismatch or perturbation; When γ EKF When =0, the new information is severely abnormal, the EKF diverges, and it is unreliable; EH∞ confidence level γ EH∞ The calculation process is as follows: Based on the sliding window residual RMS, the confidence level of EH∞ is: (82) Where, δ H∞,max The maximum allowable residual RMS threshold is obtained through offline calibration, for example, by taking the observed residual corresponding to 5% of the rated current.

[0074] Specifically, heterogeneous fusion of dual observers: Based on two confidence levels, the following fusion rule is used to select the final observer output. and : Rule 1: Normally EKF-dominated: (83) Rule 2: EH∞ Arbitration, EKF is unreliable: (84) Rule 3: Weighted fusion in the transition zone: (85) Rule 4: Dual observers are untrustworthy. (86) Simultaneously output the following observer status flag, Fobs: 00: Indicates that both observers are functioning normally; 01: Indicates low EKF confidence level, use EH∞; 10: Indicates low confidence level of EH∞, use EKF; 11: Indicates that both observers are untrusted, triggering safe mode.

[0075] The heterogeneous fusion process of dual observers is as follows Figure 4 As shown.

[0076] Specifically, a full-speed domain weight switching strategy based on an improved differential evolution algorithm.

[0077] Traditional linear weighting ignores the nonlinear changes in reconstruction error of the two methods at various speeds. In this embodiment of the invention, an improved differential evolution algorithm is used to optimize the nonlinear weighting coefficients. Based on the current speed range, the validity and weight of source C and source B are corrected. The corrected three-source values ​​will be input into the safety voting module.

[0078] (1) Division of speed range

[0079] The speed range of the power source PMSM during the flap's corresponding actions is shown in Table 1.

[0080] Table 1. Division of Full Speed ​​Range

[0081] The speed range can also be adjusted according to the actual system, such as by dividing it as a percentage of the rated speed.

[0082] (2) Improve the differential evolution algorithm to optimize the weight coefficients

[0083] Differential evolution (DE) algorithms evolve a population through mutation, crossover, and selection operations. Standard DE is sensitive to the mutation factor F and the crossover probability CR, and is prone to getting trapped in local optima. Therefore, this embodiment of the invention employs parameter adaptation, dynamically adjusting F and CR in each generation based on the success rate.

[0084] Specifically, the improved differential evolution algorithm adopts parameter adaptation, dynamically adjusting the mutation factor F and crossover probability CR in each generation based on the success rate, as follows: Step 1: Initialization NP individuals (NP=50) are uniformly generated in the λ1 space, where λ1∈[0,1] and λ2=1-λ1.

[0085] (87)

[0086] Step 2: Mutation Operation

[0087] For the t-th generation individual Generate mutated individuals: (88) Where r1, r2, and r3 are distinct random indices. The scaling factor F is adaptively adjusted. (89) Step 3: Cross-operation CR inherits components from mutated individuals with probability: (90) Crossover probability CR adaptive adjustment: (91) Step 4: Select Operation Based on the fitness function, the best candidate is selected: (92) The fitness function minimizes the sum of squared weighted speed prediction errors: (93) Where M is the number of different speed points within the switching speed domain, ω true,j The actual rotational speed (obtained through offline calibration of a high-precision encoder).

[0088] Step 5: Parameter Update

[0089] Update the next generation : (94) (95) Through offline optimization, the optimal weighting coefficients such as λ1(w) and λ2(w) at different speeds are obtained and fitted into nonlinear function curves, such as Sigmoid type or fourth-order polynomial, for online table lookup.

[0090] (3) Full-speed domain correction output

[0091] Within the switching speed domain, weighted fusion is employed: (96) (97) The full-speed domain variable weight switching process is as follows: Figure 5 As shown.

[0092] The final motor rotation angle and speed obtained by full-speed domain fusion are used as feedback signals for flap position control. At the same time, according to the motion relationship of equation (8), the deflection angle of the flap can be obtained, thus completing the full-speed domain position tracking control of the flap.

[0093] This invention establishes a forward transmission model from motor rotation angle to flap deflection angle. This model comprehensively considers three physical parameters: reduction ratio, lead of roller screw, and effective lever arm of connecting rod. The motor rotation angle obtained by state reconstruction is substituted into this model to calculate the flap deflection angle, providing physical constraints and mathematical basis for closed-loop control of flap position.

[0094] By injecting a high-frequency square wave voltage into the estimated rotating coordinate system and employing filterless current separation technology, the high-frequency response current is extracted through algebraic operations on three adjacent sampling points, eliminating the need for low-pass or band-pass filters and avoiding phase delay and amplitude attenuation caused by filters. This meets the requirements for dynamic position reconstruction performance of aircraft flaps under low-speed fine-tuning positioning and zero-speed holding conditions. The position error signal is extracted using the vector cross multiplication method, avoiding the sensitivity of arctangent operations to noise and enhancing the reconstruction robustness in strong electromagnetic interference environments. The rotor position and speed are reconstructed using an improved Romberg observer that incorporates load torque as an extended state, effectively compensating for reconstruction errors caused by sudden changes in aerodynamic loads during flight.

[0095] A four-dimensional state-space model is established using extended Kalman filtering, outputting the state reconstruction and chi-square statistics of the innovation to ensure high dynamic response performance under normal flight conditions. A six-dimensional augmented state-space model is established using extended H∞ filtering, outputting the state reconstruction and root mean square residuals to ensure robustness when model parameter perturbations or noise statistics are unknown. The confidence scores of EKF and H∞ are calculated to achieve a quantitative assessment of the reliability of the two observers. The fusion results of the two observers are output according to the fusion rules to achieve complementary fusion of EKF and EH∞. The fusion rules include EKF-dominated mode, H∞-dominated mode, weighted fusion mode, and unusable degradation processing mode to ensure the continuity and reliability of position reconstruction within the entire flight envelope.

[0096] By constructing three independent state information sources—physical sensors, heterogeneous fusion output from dual observers, and improved high-frequency square wave injection observations—a heterogeneous redundancy architecture combining physical and analytical methods is formed to address potential sensor failures such as wire breakage, jamming, and offset under strong vibrations and wide temperature ranges in aviation environments. The dual observer fusion output is generated through confidence assessment and weighted fusion using EKF and H∞ dual observers, while the improved high-frequency injection observations are generated through filterless current separation and an improved Romberg observer. The multi-source signal processing operates in parallel and independently, without dependency on each other, providing reliable position feedback even when a single information source fails, thus meeting the requirement that the probability of catastrophic failure in aviation flap control systems is less than 1e. -9 Safety requirements per flight hour.

[0097] In this embodiment of the invention, the operating range is divided into three regions based on the current speed reconstruction value: the zero-low speed region, the transition speed region, and the medium-high speed region. Multiple position sources are weighted and fused. The zero-low speed region mainly uses the improved high-frequency injected observation value, the medium-high speed region mainly uses the fused output of the dual observers, and the weight coefficients of the transition speed region are optimized offline by an improved differential evolution algorithm and fitted into a nonlinear curve for online lookup. This achieves smooth coverage of the full speed range state reconstruction of the flap PMSM and completes the closed-loop control of the flap position.

[0098] In one specific embodiment of the present invention, an aircraft flap PMSM multi-source heterogeneous fusion position information reconstruction system includes: The physical layer is used to acquire multi-source aviation signals and preprocess the aviation signals to generate standardized data. The analytical redundancy layer is used to perform parallel and independent computation on the standardized data using a heterogeneous redundancy architecture that combines physical and analytical methods to obtain multi-source redundant signals. The full-speed domain fusion decision layer is used to divide different operating conditions according to the complete flight mission profile of the flaps, perform condition-adaptive weighted fusion of the multi-source redundant signals, and output the reconstructed values ​​of the motor position and speed. The output layer is used to call the pre-established kinematic model of the flap electromechanical actuator, solve the reconstructed position and speed values ​​to obtain the actual deflection angle of the flap, and output the final motor rotation angle, speed and flap angle.

[0099] Specifically, such as Figure 1 As shown, the first layer, the physical layer, is responsible for aviation signal acquisition, including a rotary transformer (source A), current and voltage sensors, and a PWM drive protection circuit that meets the characteristics of aviation power supplies. After Clarke transformation and preprocessing, the raw signals are converted into unified coordinate components, providing a data foundation for the upper-layer observation modules.

[0100] The second layer: an analytical redundancy layer enables parallel and independent computation of the three sources. Source A consists of values ​​directly acquired by physical sensors; source C employs an improved high-frequency square wave injection method, extracting position errors through filterless current separation and vector cross multiplication, combined with an improved Romberg observer incorporating load torque estimation, to output high-precision position reconstruction under low-speed flap fine-tuning positioning and zero-speed holding conditions; source B comprises dual observers, EKF and EH∞, achieving complementary high dynamic range under normal conditions and strong robustness under abnormal conditions through confidence assessment and heterogeneous fusion. The three sources are independent of each other, forming a heterogeneous redundancy architecture that combines physical and analytical methods.

[0101] The third layer: The full-speed domain fusion decision layer divides the flap full-flight mission profile into three operating condition intervals. The zero-low speed domain is determined based on the observation results of source C, and the medium-high speed domain is determined based on the observation results of source B. The zero-low speed domain is dominated by source C, and the medium-high speed domain is dominated by source B. The transition speed domain is weighted and fused using nonlinear weights optimized offline by an improved differential evolution algorithm, outputting the final position and rotational speed reconstruction values ​​to ensure the continuity and reliability of flap position reconstruction within the full-flight mission profile.

[0102] The fourth layer: The output layer is based on the kinematic model of the flap electromechanical actuator. It calculates the actual deflection angle of the flap by reconstructing the motor rotation angle, and outputs the motor rotation angle, speed and flap angle to the three closed-loop control system to meet the flap angle requirements of each stage of takeoff, cruise and landing.

[0103] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0104] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for reconstructing the position information of an aircraft flap PMSM from multiple sources heterogeneous fusion, characterized in that, include: Collect multi-source aviation signals and preprocess the aviation signals to generate standardized data; A heterogeneous redundancy architecture combining physical and analytical methods is used to perform parallel and independent computation on the standardized data to obtain multi-source redundant signals; Different operating conditions are divided according to the complete flight mission profile of the flaps. The multi-source redundant signals are subjected to condition-adaptive weighted fusion to output the reconstructed values ​​of the motor position and speed. The pre-established kinematic model of the flap electromechanical actuator is called to solve the reconstructed position and speed values ​​to obtain the actual deflection angle of the flap, and the final motor rotation angle, speed and flap angle are output.

2. The method for reconstructing the position information of an aircraft flap PMSM multi-source heterogeneous fusion according to claim 1, characterized in that, A heterogeneous redundancy architecture combining physical and analytical methods is used to perform parallel and independent computations on the standardized data, including: Source A is the measured rotor position signal directly acquired by a physical sensor; Source B employs dual observers, EKF and EH∞, and outputs medium-to-high-speed analytical position observation signals through confidence assessment and heterogeneous fusion. Source C employs an improved high-frequency square wave injection method, extracts position errors through filterless current separation and vector cross multiplication, and combines an improved Romberg observer that incorporates load torque estimation to output a zero-low-speed analytical position observation signal.

3. The method for reconstructing the position information of an aircraft flap PMSM multi-source heterogeneous fusion according to claim 2, characterized in that, The operating conditions include the zero-low speed domain, the medium-high speed domain, and the transition speed domain. The zero-low speed domain is determined based on the observation results of source C, the medium-high speed domain is determined based on the observation results of source B, and the transition speed domain is weighted and fused by nonlinear weights optimized offline by an improved differential evolution algorithm to output the final position and speed reconstruction values.

4. The method for reconstructing the position information of an aircraft flap PMSM multi-source heterogeneous fusion according to claim 2, characterized in that, Source C employs an improved high-frequency square wave injection method, extracting position errors through filterless current separation and vector cross product. Combined with an improved Romberg observer incorporating load torque estimation, it reconstructs the output position under low-speed flap fine-tuning and zero-speed holding conditions, including: A high-frequency model of a built-in permanent magnet synchronous motor under high-frequency excitation is established by alternately injecting constant-amplitude high-frequency square wave excitation voltage into the rotating coordinate system of the motor. By utilizing the fundamental current characteristics at three adjacent discrete sampling times, the high-frequency response current along the α-β axis is obtained by solving the current equations at the three sampling points simultaneously. Based on the high-frequency model, under square wave injection, the current change in each sampling period is calculated to obtain the high-frequency incremental current; The error signal is extracted from the separated high-frequency incremental current using the vector cross product method; Rotor reconfiguration position, reconfiguration speed, and load torque are selected as state variables. A Romberg observer based on load disturbance compensation is constructed, and the observation state is corrected according to the error signal to output the rotor position and speed observation values ​​under zero low-speed conditions.

5. The method for reconstructing the position information of an aircraft flap PMSM multi-source heterogeneous fusion according to claim 4, characterized in that, A high-frequency model of a built-in permanent magnet synchronous motor under high-frequency excitation is established by alternately injecting constant-amplitude high-frequency square wave excitation voltages into the rotating coordinate system of the motor, including: The voltage equation of the built-in permanent magnet synchronous motor in the dq rotating coordinate system is: ； Where p = d / dt, R s L is the stator resistance. d , L q These are the inductances along the d and q axes, respectively, where ω is the electric angular velocity and ψ is f For permanent magnet flux linkage; The high-frequency voltage equation is expressed as: ； Give the estimated rotating coordinate system Injecting high-frequency square wave voltage: ； Among them, V h Let (-1) represent the magnitude of the injected voltage. k The alternation of signs corresponds to the injection frequency. f h = f PWM / 2, sampling period T s =1 / f PWM ; By transforming the coordinates, we obtain the high-frequency model in the stationary a-β coordinate system: ； in, ;θ e This is the actual rotor electrical angle. The reconstructed rotor electrical angle, reconstruction error .

6. The method for reconstructing the position information of an aircraft flap PMSM multi-source heterogeneous fusion according to claim 2, characterized in that, Employing a dual observer architecture of EKF and EH∞, complementary high dynamics under normal conditions and strong robustness against anomalies are achieved through confidence assessment and heterogeneous fusion, including: A PMSM model is constructed in the α-β stationary coordinate system. The process noise and measurement noise are assumed to follow a Gaussian distribution. Prediction-update iterations are completed through Euler discretization and Jacobian matrix solving. A chi-square consistency test is constructed based on the observed information sequence, and the EKF confidence level γ is quantitatively calculated. EKF Output the rotor position observation results under normal flight conditions; The back EMF, inductance, and resistance are simultaneously reconstructed as state variables to build an augmented state model. The H∞ filtering cost function is iteratively solved, and the rotor position is extracted from the augmented state. The EH∞ confidence level γ is calculated using root mean square quantization of the residuals through a sliding window. EH∞ Output the rotor position observation results under motor parameter perturbation and electromagnetic interference conditions; Based on EKF confidence level γ EKF and EH∞ confidence level γ EH∞ Four working modes are defined: EKF normal dominant mode, EH∞ arbitration mode, transition zone weighted fusion mode, and dual observer untrusted mode. The observer fault status flag is output synchronously, and the single-channel medium and high speed rotor position observation value is fused and output.

7. The method for reconstructing the position information of an aircraft flap PMSM through multi-source heterogeneous fusion according to claim 6, characterized in that, EKF confidence level γ EKF The calculation process is as follows: Based on chi-square statistics To define the EKF confidence level: ； EH∞ confidence level γ EH∞ The calculation process is as follows: Based on the sliding window residual RMS, the confidence level of EH∞ is: ； Where, δ H∞,max The maximum allowed residual RMS threshold.

8. The method for reconstructing the position information of an aircraft flap PMSM multi-source heterogeneous fusion according to claim 3, characterized in that, The improved differential evolution algorithm employs parameter adaptation, dynamically adjusting the mutation factor F and crossover probability CR in each generation based on the success rate. The process is as follows: NP individuals are uniformly generated in the λ1 space, where λ1∈[0,1] and λ2=1-λ1; ； For the t-th generation individual Generate mutated individuals: ； Where r1, r2, and r3 are distinct random indices, and the scaling factor F is adaptively adjusted: ； Inherited components from mutated individuals using crossover probability CR: ； Crossover probability CR adaptive adjustment: ； Based on the fitness function, the best candidate is selected: ； The fitness function minimizes the sum of squared weighted speed prediction errors: ； Where M is the number of different speed points within the switching speed domain, ω true,j This refers to the actual rotational speed; Update the next generation : ； ； The optimal weighting coefficients for different speeds were obtained through offline optimization.

9. A multi-source heterogeneous fusion position information reconstruction system for an aircraft flap PMSM, employing the multi-source heterogeneous fusion position information reconstruction method for an aircraft flap PMSM as described in any one of claims 1-8, characterized in that, include: The physical layer is used to acquire multi-source aviation signals and preprocess the aviation signals to generate standardized data. The analytical redundancy layer is used to perform parallel and independent computation on the standardized data using a heterogeneous redundancy architecture that combines physical and analytical methods to obtain multi-source redundant signals. The full-speed domain fusion decision layer is used to divide different operating conditions according to the complete flight mission profile of the flaps, perform condition-adaptive weighted fusion of the multi-source redundant signals, and output the reconstructed values ​​of the motor position and speed. The output layer is used to call the pre-established kinematic model of the flap electromechanical actuator, solve the reconstructed position and speed values ​​to obtain the actual deflection angle of the flap, and output the final motor rotation angle, speed and flap angle.

Images