A method for multi-parameter synchronous testing and performance evaluation of a drone power system

Abstract

The present application relates to the technical field of unmanned aerial vehicle power system test evaluation, in particular to a kind of unmanned aerial vehicle power system multi-parameter synchronous test and performance evaluation method, comprising: establishing typical flight mission profile library and converting into dynamic throttle command sequence;Build hardware synchronous test bench based on Beidou timing and FPGA, collect thrust, torque, speed, voltage and current, multi-point temperature and vibration in sub-microsecond level time;After adaptive filtering and interpolation alignment to synchronous signal, use current ripple to extract sensorless speed, combine extended Kalman filter to jointly estimate motor electrical and thermal parameters, decouple calculation electric governor, motor, propeller dynamic efficiency and thrust transient response, thermal decay and vibration characteristics;Construct four-dimensional evaluation matrix of power performance, energy efficiency, thermal stability and vibration quality, dynamically adjust weight according to task segment using combination weighting, calculate comprehensive closeness by grey correlation degree, output performance level and multi-dimensional radar chart.The present application realizes dynamic energy chain cross decoupling and multi-dimensional evaluation of mission adaptation, significantly improves the authenticity and comprehensiveness of test.

Description

Technical Field

[0001] This invention relates to the field of testing and evaluation technology for unmanned aerial vehicle (UAV) power systems, and in particular to a method for simultaneous testing and performance evaluation of multiple parameters of UAV power systems. Background Technology

[0002] The performance of an electric drone's power system directly determines its endurance, maneuverability, safety margins, and life-cycle cost. The power system consists of an electronic speed controller, a brushless motor, and a propeller; these three components are strongly coupled and experience complex and transient conditions during real flight, ranging from hovering to rapid acceleration, climb, cruise, and gust disturbances. Currently used testing and evaluation methods have the following significant shortcomings: Lack of dynamic decoupling: Most existing dynamometers can only collect steady-state tension, torque, speed, and electrical power, and cannot simulate the rapid changes in UAV flight control commands. Even if some solutions can provide ramp or step signals, they lack synchronous high-frequency recording of parameters such as current, temperature, and vibration. This makes it impossible to separate the individual contributions of ESC efficiency, motor electromagnetic efficiency, and propeller aerodynamic efficiency under dynamic operating conditions, and to pinpoint whether the efficiency bottleneck lies in power conversion, electromechanical energy conversion, or aerodynamic propulsion.

[0003] Inconsistent time bases for multiple parameters: Various sensors, such as those for tension, torque, voltage, current, temperature, and vibration, often operate independently, each with its own internal clock, making post-processing alignment difficult. Especially within a tens of millisecond window where current spikes, torque pulsations, and vibration shocks occur simultaneously, time deviations can cause serious errors in energy flow analysis, even leading to results that violate physical laws. Current technologies have not resolved the issue of sub-microsecond synchronous triggering and unified data packet timestamps at the hardware level.

[0004] The evaluation dimensions are singular and static: evaluation indicators are mostly limited to maximum thrust and hovering efficiency, failing to consider thrust response speed, overshoot, thermal decay characteristics, vibration intensity, and its frequency components. Furthermore, fixed weights cannot reflect the differences in power system requirements across different flight mission phases—hovering emphasizes efficiency and thermal stability, rapid acceleration emphasizes response bandwidth and overshoot suppression, while cruise focuses on system resonance avoidance. Ignoring mission context results in a lack of specificity and foresight in selection and health assessment.

[0005] Limited parameter identification methods: Key parameters such as motor winding temperature, permanent magnet flux linkage, and inductance cannot usually be directly measured by installing sensors. Traditional methods require offline identification and cannot reflect parameter drift caused by temperature changes and demagnetization effects during flight, thus making it impossible to correct the efficiency decoupling model online.

[0006] In summary, there is an urgent need for an innovative method that can inject real flight mission profiles into the testing process, achieve microsecond-level aligned acquisition of multiple physical quantities such as electricity, force, heat, and vibration through hardware synchronization, jointly estimate the internal state of the motor using the characteristics of electrical signals, dynamically decouple the ESC-motor-propeller three links, and finally complete the comprehensive performance quantitative evaluation with mission adaptive weights. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for multi-parameter synchronous testing and performance evaluation of unmanned aerial vehicle power systems, so as to solve the problems existing in the background art.

[0008] This invention provides the following technical solution: a method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system, comprising the following steps: S1. Establish a typical flight mission profile library and convert it into dynamic throttle command sequences; S2. Build a multi-physical quantity synchronous test bench, and unify the acquisition clocks of thrust, torque, speed, voltage and current, multi-point temperature and vibration sensors to the same reference through a hardware synchronous triggering network based on satellite time synchronization and FPGA. S3. The real-time controller executes a dynamic command sequence with a fixed step size to drive the tested ESC-motor-propeller system and synchronously triggers all sensors to collect data, thereby obtaining a multi-parameter timing signal with a unified timestamp. S4. Adaptive filtering and interpolation alignment are performed on the multi-parameter time-series signal. The sensorless rotation speed is extracted using the current ripple. The electrical parameters and rotor temperature of the motor are estimated in conjunction with the extended Kalman filter. Then, the dynamic efficiency of the electronic speed controller, the electromagnetic efficiency and mechanical efficiency of the motor, and the aerodynamic efficiency of the propeller are decoupled and calculated. At the same time, the thrust transient response index and the temperature-thrust attenuation coefficient are extracted. S5. Extract the vibration level and characteristic frequency amplitude across the entire vibration band, and establish a four-dimensional evaluation matrix that includes dynamic performance, energy efficiency, thermal stability, and vibration quality; S6. Based on the task segment type to which the current dynamic instruction belongs, the weights of each dimension are adjusted in real time using a combined weighting method. The overall performance closeness is calculated by using grey relational analysis or the sorting method to approximate the ideal solution, and the performance level and multi-dimensional radar chart are output.

[0009] By establishing a flight mission profile library and converting it into dynamic command sequences, the test conditions are extended from steady state to the transient process of real flight, making up for the shortcomings of traditional methods that can only evaluate static performance. A hardware synchronous trigger bus is used to collect data on multiple physical quantities in a timely manner, completely solving the problem of distortion in dynamic energy flow analysis caused by inconsistent time bases of multiple sensors. Current ripple is used to extract rotational speed, and combined with a state observer to identify electrical and thermal parameters, achieving for the first time precise decoupling of the efficiency of the ESC, motor, and propeller under dynamic conditions, allowing for the location of efficiency bottleneck components. The transient response of thrust and the amplitude of vibration characteristic frequencies are extracted to construct a four-dimensional evaluation matrix, breaking through the previous single evaluation mode that only focused on thrust and efficiency. Based on real-time adjustment of weights according to mission segments, combined weighting and grey relational evaluation are performed, ensuring that the evaluation results are highly matched with flight mission requirements, upgrading from "whether it can be used" to "how well it performs in which mission," providing a direct basis for customized mission selection and predictive maintenance.

[0010] Furthermore, in step S2, the hardware synchronization triggering network uses the second pulse and 10MHz clock output by the Beidou / GPS satellite receiving module as the system clock source, uses the FPGA to generate multi-channel synchronization triggering pulses, and adds a timestamp containing absolute time and sampling sequence number to each sampling point, so that the sampling time deviation between different sensor channels is less than 50 nanoseconds.

[0011] By combining BeiDou / GPS timing with FPGA logic to generate synchronization pulses, the synchronization accuracy is better than 100 nanoseconds. This enables sensors with different physical properties such as tension, torque, current, temperature, and vibration to achieve sub-microsecond alignment at the sampling time, fundamentally ensuring the time accuracy of subsequent cross-decoupling and energy flow calculation, and avoiding efficiency calculation errors and transient characteristic misalignment caused by time deviation.

[0012] Furthermore, in step S4, the specific method for extracting sensorless rotational speed using current ripple is as follows: The electronically controlled DC bus current or three-phase AC current is collected, the commutation frequency is extracted from it, and the instantaneous rotational speed is obtained through a phase-locked loop or zero-crossing detection. Then Mechanical speed obtained from torque sensor Fusion, as an observation of the extended Kalman filter.

[0013] Instantaneous speed is extracted by utilizing the high-frequency commutation ripple in the motor phase current, eliminating the need for an additional speed sensor and reducing system complexity and cost. By inputting the speed and torque sensor signals together into an extended Kalman filter, internal parameters that are difficult to measure directly, such as rotor temperature, winding inductance, and flux linkage, can be estimated online. Based on this, copper loss, iron loss, and mechanical loss can be separated, and the efficiency decoupling of each component is based on real-time and accurate physical parameters, which is significantly better than traditional methods that use offline constants or empirical formulas.

[0014] Furthermore, the state vector estimated by the extended Kalman filter includes at least rotor speed, winding inductance, permanent magnet flux linkage, and rotor temperature, with voltage, current, torque, and mechanical speed as observations, and outputs the motor electromagnetic torque, various loss components, and efficiency values ​​in real time.

[0015] Furthermore, the thrust transient response indicators extracted in step S4 include the step response rise time t. 90 Overshoot σ, settling time, and steady-state error; the temperature-thrust attenuation coefficient α is defined as: Where ΔT is the stator temperature rise. and These represent the thrust before and after the temperature rise, respectively.

[0016] The thrust step response index is quantified into rise time, overshoot, steady-state error, and thrust thermal decay coefficient after temperature compensation. This comprehensively characterizes the transient tracking capability of the power system under rapid command changes and its performance retention capability after temperature rise, providing a quantifiable basis for evaluating the power adaptability of UAVs in harsh scenarios such as rapid acceleration and gust compensation.

[0017] Furthermore, in step S5, the vibration characteristic frequencies include at least the 1P and 2P of the corresponding rotational speed fundamental frequency and the blade passing frequency BPF; the vibration quality index includes the full-band vibration level in the 10~1000Hz frequency range and the amplitude of the above characteristic frequencies.

[0018] Furthermore, in step S6, the task segment types include hovering, climbing, cruising, rapid maneuvering, and tilt transition. For each task segment, a subjective weight vector obtained by the analytic hierarchy process is preset, and then the objective weight is determined based on the current amount of data information using the entropy weight method. The final weight is obtained by multiplication synthesis. The calculation of the overall performance closeness adopts the grey relational analysis method.

[0019] By employing a combination of entropy weighting and hierarchical analysis for weighting, and automatically switching weight vectors based on mission types such as hovering, climbing, cruise, and gust disturbances, the evaluation results can adapt to the performance requirements of different flight phases. This avoids the drawbacks of fixed-weight evaluations that are "one-size-fits-all" in complex flight missions, and significantly improves the relevance and practicality of the evaluation.

[0020] Furthermore, after step S4, a health warning step is also included: continuously monitoring the residual between electromagnetic torque and mechanical torque, as well as the sideband modulation of the vibration characteristic frequency. When the residual exceeds the dynamic threshold or a specific fault characteristic appears, demagnetization, inter-turn short circuit, or bearing wear fault is determined and associated with the specific action in the task profile.

[0021] After decoupling the calculation, a fault warning is introduced. By monitoring the residuals of electromagnetic torque and mechanical torque and vibration sidefrequency characteristics, early faults such as demagnetization, winding faults or bearing damage can be identified simultaneously during performance testing. The faults can be accurately located to specific task actions and time points, combining testing and health management functions into one, thereby improving the safety assurance capability of the system throughout its entire life cycle.

[0022] The technical effects and advantages of this invention are as follows: By using mission profile injection and a BeiDou / FPGA-based synchronous triggering network, microsecond-level aligned acquisition of current spikes, torque pulsations, and vibration shocks within the flight envelope was achieved for the first time in the laboratory, solving the fundamental problem of dynamic energy flow measurement distortion.

[0023] By utilizing current ripple sensorless speed measurement and extended Kalman filtering, the electrical and thermal parameters of the motor are estimated in real time during transient processes. The dynamic efficiency curves of the ESC, motor, and propeller are accurately decoupled, and bottleneck components are located, filling the technical gap in dynamic efficiency separation in this field.

[0024] A comprehensive evaluation system covering power performance, energy efficiency, thermal stability and vibration quality was constructed, and the weights were dynamically adjusted according to the mission segment, so that the evaluation results were upgraded from "whether it can fly" to "which mission it performs best in", providing a direct basis for mission-customized selection and predictive maintenance.

[0025] By utilizing synchronous multi-parameter residual analysis and vibration sideband characteristics, early detection of minor faults can be achieved while evaluating performance, and these faults can be accurately mapped to induced actions in the mission profile, significantly improving flight safety and availability. Attached Figure Description

[0026] Figure 1 This is a flowchart of a method for synchronous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to the present invention. Figure 2 This is a flowchart illustrating the construction of the energy chain cross-decoupling and multi-dimensional evaluation index system of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0028] Please see Figure 1 As shown, a method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) propulsion system includes the following steps: S1. Establish a library of typical flight mission profiles and convert them into dynamic throttle command sequences. In practice, for the type of UAV under test, collect or generate typical flight mission data through flight simulation. The mission types should include at least hovering, vertical climb, horizontal acceleration, cruise, gust disturbances, and emergency go-around. Using the UAV's kinematics and aerodynamic models, the altitude, attitude, and velocity changes of each mission profile are inversely solved into a dynamic sequence of the throttle PWM duty cycle required by the power system over time. To ensure that dynamic characteristics are not lost, the command update rate is not less than 500Hz, and a small high-frequency sinusoidal perturbation is superimposed on the command to excite the high-frequency response of the system. Finally, a set of digital command files equivalent to the actual flight power spectrum is formed and stored in the real-time controller. S2. Construct a multi-physical quantity synchronous test bench. Through a hardware synchronous triggering network based on satellite timing and FPGA, unify the acquisition clocks of thrust, torque, speed, voltage, current, multi-point temperature, and vibration sensors to a single reference. In specific implementation, the tested objects include an electronic speed controller, a brushless DC motor, and a propeller. The test bench base is a rigid platform made of cast iron or marble, on which a six-axis force / torque sensor is mounted. The motor is fixed to the sensor via a dedicated mounting bracket to simultaneously measure axial thrust, counter-torque, and radial unbalanced force. A high-bandwidth dynamic torque-speed composite sensor is connected in series between the motor output shaft and the propeller to measure shaft torque and speed in real time. The DC input terminal of the electronic speed controller is equipped with Hall effect voltage and current sensors with a sampling bandwidth of not less than 200kHz to capture current ripple; for more precise identification, a current sensor can also be added to the three-phase output terminal of the electronic speed controller. Temperature measurement employs at least three NTC thermistors or K-type thermocouples, respectively attached to the motor stator core, winding ends, and the heatsink of the electronically controlled MOSFET. An infrared thermal imager is also used to monitor the overall temperature field. Vibration measurement utilizes a laser Doppler vibration meter, with its laser spot aligned radially with the motor housing to acquire vibration velocity signals non-contactly.

[0029] All the aforementioned sensor signals are connected to a synchronization control unit. This unit integrates a BeiDou / GPS dual-mode timing module and an FPGA chip. The BeiDou / GPS module provides a PPS second pulse accurate to 1 second and a 10MHz temperature-controlled crystal oscillator reference clock; the FPGA uses this clock to generate multiple synchronization trigger pulses, which are sent to the signal conditioners or acquisition cards of each sensor, while simultaneously adding a standardized timestamp containing absolute time (seconds, nanoseconds) and sampling sequence number to each sampling point. All channels are aggregated to the host computer through an industrial Ethernet switch supporting the IEEE 1588 precision time protocol. Actual measurements show that the sampling time deviation of each channel is less than 50 nanoseconds, ensuring strict synchronization of multiple parameters. S3. The real-time controller executes a dynamic command sequence in fixed steps to drive the tested ESC-motor-propeller system and synchronously triggers all sensors to collect data, obtaining a multi-parameter timing signal with a unified timestamp. In specific implementation, the real-time controller reads the duty cycle value from the command file in fixed steps of 0.1 milliseconds and converts it into a PWM signal to drive the electronic speed controller. At the beginning of each step, the synchronous control unit broadcasts a trigger signal, causing all sensors to latch and upload the current data at the same time. The entire mission profile is continuously collected, forming a raw data matrix with multiple columns and strict time alignment, including thrust, torque, speed, voltage, current, multi-point temperature, and triaxial vibration signals. S4. Adaptive filtering and interpolation alignment are performed on the multi-parameter time-series signal. The sensorless rotation speed is extracted using the current ripple. The electrical parameters and rotor temperature of the motor are estimated in conjunction with the extended Kalman filter. Then, the dynamic efficiency of the electronic speed controller, the electromagnetic efficiency and mechanical efficiency of the motor, and the aerodynamic efficiency of the propeller are decoupled and calculated. At the same time, the thrust transient response index and the temperature-thrust attenuation coefficient are extracted. In practical implementation, firstly, variational mode decomposition (VMD) is used to adaptively filter the original signal. This decomposes the signal into eigenmode functions with different center frequencies, eliminating high-frequency noise and electromagnetic interference components while preserving transient characteristics such as step and impulse signals. To address the issue of varying original sensor sampling rates, low-sampling-rate signals are resampled using cubic spline interpolation, uniformly interpolated to the time grid of the highest sampling rate, ensuring that the alignment error is less than one sampling period.

[0030] Next, the sensorless rotational speed is extracted from the filtered current signal. For DC bus current, the fundamental frequency of the ripple caused by commutation is extracted using a sliding Goertzel algorithm or a phase-locked loop-based tracking filter. For three-phase current, the current vector angular frequency can be obtained through the Clark-Parker transform. The obtained ripple rotational speed is then... Mechanical speed directly measured by torque sensor Compare and integrate.

[0031] Construct an extended Kalman filter for joint estimation of the motor state. The state vector is taken as... Where ω is the rotor electrical angular velocity and θ is the rotor position angle. , For the direct-axis current, Stator phase resistance, , It is a direct-axis inductor. It is a permanent magnet flux chain. Let be the rotor temperature. The system state equations are the voltage equations and thermal model of the motor in the dq coordinate system, and the observation equations are... , The system measures electromagnetic torque (obtained from a torque sensor) and rotational speed (a fusion of current ripple and mechanical speed). Through iterative prediction and updating, it outputs various electrical parameters and rotor temperature in real time. These parameters allow for the precise calculation of the motor's copper losses, iron losses, and mechanical losses.

[0032] Based on this, the dynamic efficiency of each component is decoupled. Electronic speed controller efficiency: in Based on MOSFET switching and conduction loss models, and combined with real-time current and case temperature estimations, the motor electromagnetic efficiency is as follows: electromagnetic torque , where p is the extreme logarithm.

[0033] Motor mechanical efficiency: This is the measured value from the torque sensor.

[0034] Propeller aerodynamic efficiency: Where T is the thrust. The induced velocity is obtained by iteratively solving for thrust, air density, and propeller disk area based on momentum theory. Instantaneous curves of all efficiency values ​​versus time are obtained, and the energy-weighted average efficiency of the entire mission profile can be calculated.

[0035] Simultaneously, dynamic response indices are extracted from the thrust step response curve: the time required for the thrust to rise from 10% to 90% of its final value is defined as the rise time t. 90 The overshoot σ is the percentage of the difference between peak thrust and steady-state thrust relative to the steady-state thrust. The settling time is the time it takes for the thrust to enter and remain within the ±2% error band of the steady-state value, and the steady-state error is recorded. The temperature-thrust attenuation coefficient is calculated by measuring the thrust under different temperature conditions. in and The thrust in cold and hot states, respectively. and This corresponds to the stator temperature. The thermal time constant is obtained by fitting the temperature rise curve.

[0036] Vibration signal processing: Short-time Fourier transform (window length 512 points, 75% overlap) was performed on the vibration velocity to obtain the time spectrum. The root mean square vibration level (RMS) in the frequency range of 10Hz to 1000Hz was extracted (reference value 1μm / s). Simultaneously, order tracking was performed using the rotational speed signal to identify the vibration amplitude at 1P (fundamental rotational speed frequency), 2P (double rotational speed frequency), and BPF (blade passage frequency = rotational speed × number of blades).

[0037] S5. Extract the vibration level and characteristic frequency amplitude across the entire vibration band, and establish a four-dimensional evaluation matrix that includes dynamic performance, energy efficiency, thermal stability, and vibration quality; In practice, all the above parameters are categorized into four primary indicator dimensions, with each dimension having several secondary indicators: Dynamics: Maximum thrust, maximum torque, thrust step rise time t 90 Overshoot σ, settling time; Energy efficiency: hovering power efficiency (g / W), cruise overall efficiency, electronic governor efficiency, motor electromagnetic efficiency, mechanical efficiency, propeller efficiency, system efficiency MAP; Thermal stability: steady-state temperature rise under rated thrust, thermal time constant, temperature-thrust attenuation coefficient α; Vibration quality: full-band vibration level, 1P amplitude, 2P amplitude, BPF amplitude, characteristic frequency sideband modulation index.

[0038] Dimensionless normalization is performed on each secondary indicator. For positive indicators (the larger the better, such as efficiency), the following method is adopted: For negative indicators (the smaller the better, such as overshoot), the following approach is adopted: For intermediate indicators, a distance function is constructed based on the target value. All indicator values ​​are linearly mapped to the [0,1] interval to obtain a normalized evaluation matrix; S6. Based on the task segment type to which the current dynamic instruction belongs, the weights of each dimension are adjusted in real time using the combined weighting method. The overall performance closeness is calculated by the grey relational degree or the sorting method that approximates the ideal solution, and the performance level and multi-dimensional radar chart are output. In practice, the mission segment type (hovering, climb, cruise, rapid maneuver, etc.) is automatically identified based on the current flight phase of the dynamic command sequence. The identification method is as follows: a simple rule classifier is set based on thrust, throttle change rate, and flight attitude reference signals. For example, if the average throttle value is near the hovering value and the change rate is below the threshold, it is determined to be hovering; if the throttle continues to increase and the thrust increases, it is determined to be climbing, and so on.

[0039] For each task segment, a subjective weight vector is pre-constructed using the Analytic Hierarchy Process (AHP). Domain experts are invited to conduct pairwise comparisons of the importance of dynamic performance, energy efficiency, thermal stability, and vibration quality using a 1-9 scale, forming a judgment matrix. The subjective weights W are then obtained after passing a consistency check. sub Meanwhile, for the current test data to be evaluated, the objective weight W is calculated based on the information entropy of each indicator. obj The lower the information entropy of a certain indicator, the greater the amount of information it provides, and the higher its objective weight. The combined weights are ultimately obtained using a multiplicative synthesis method. This ensures that the weights reflect both expert experience and the differences in the data itself.

[0040] The columns of the normalized evaluation matrix are weighted using corresponding combined weights to form a weighted standardized matrix. Then, grey relational analysis is used to calculate the correlation between each evaluation object (or different configurations of the same object) and the ideal optimal solution. All indicators of the ideal optimal solution are set to a maximum value of 1. The grey relational coefficient between each indicator and the ideal solution is calculated, and then the overall closeness is obtained. The higher the closeness, the better the performance. The closeness is mapped to a percentage score and divided into four levels: A (≥90), B (80-89), C (60-79), and D (<60). Simultaneously, the scores for four dimensions—dynamic performance, energy efficiency, thermal stability, and vibration quality—are output in radar chart form for easy and intuitive comparison.

[0041] Specifically, in step S2, the hardware synchronization triggering network uses the second pulse and 10MHz clock output by the Beidou / GPS satellite receiving module as the system clock source, uses the FPGA to generate multi-channel synchronization triggering pulses, and adds a timestamp containing absolute time and sampling sequence number to each sampling point. The sampling time deviation between different sensor channels is less than 50 nanoseconds.

[0042] In practical implementation, the hardware synchronization trigger network is crucial for achieving microsecond-level time alignment in the synchronization test bench described in step two. The BeiDou / GPS timing module uses a multi-mode receiver supporting timing accuracy better than 20 nanoseconds. Its output PPS (Power Pulse Per Second) and 10MHz clock signal are fed into an FPGA chip (such as the Xilinx Artix-7 series). An internal timestamp counter is built into the FPGA, accumulating using a 10MHz clock and being synchronously reset by the PPS pulse every second. The FPGA generates multiple programmable trigger pulses, which are connected to the external trigger input ports of the six-axis force sensor acquisition card, torque-speed sensor conditioning module, voltage and current acquisition module, temperature monitoring instrument, and laser vibration meter controller. When the real-time controller issues a "acquire one step" command, the FPGA immediately generates a rising edge trigger pulse and simultaneously writes the counter value (i.e., the timestamp) at that moment into the data frame header of each acquisition channel via the backplane bus. Each sensor completes sampling at the trigger edge, ensuring inter-channel synchronization accuracy within 50 nanoseconds. The host computer receives data packets with the same timestamp via industrial Ethernet (using the IEEE 1588 protocol to further synchronize the clocks of each node in the network), achieving perfect alignment of the data stream.

[0043] Specifically, in step S4, the method for extracting sensorless rotational speed using current ripple is as follows: The electronically controlled DC bus current or three-phase AC current is collected, the commutation frequency is extracted from it, and the instantaneous rotational speed is obtained through a phase-locked loop or zero-crossing detection. Then Mechanical speed obtained from torque sensor Fusion, as an observation of the extended Kalman filter.

[0044] In practical implementation, for brushless DC motors driven by electronic speed controllers, the DC bus current contains a ripple component proportional to the commutation frequency. In this embodiment, the DC bus current is sampled at a 50kHz sampling frequency, and after passing through a bandpass filter (center frequency tracking predicts the speed range), the ripple frequency f is extracted using a phase-locked loop based on a second-order generalized integrator. ripple Since a motor undergoes N commutations per revolution, where N is related to the number of pole pairs and the commutation method, the speed cannot be determined without sensors. The speed and the speed output by the torque sensor (Calculated from pulse signals) Fusion is performed: when the mechanical speed fluctuation is small, it is based on... To calibrate the N value as a reference and eliminate accumulated errors; in fast dynamics, It has higher bandwidth and can reflect instantaneous fluctuations.

[0045] The state equations of the extended Kalman filter are in discretized form: State prediction: State vector: enter Let f be the voltage across the dq axis. The nonlinear function f is composed of the motor voltage equations and the rotor thermal model, for example: It is also assumed that the parameters change slowly, such as dR_s / dt≈0. Observation vector: in for and Weighted fusion results. The observation matrix H is obtained at each time step through linearization. All states can be recursively estimated using the standard prediction-update iteration of the EKF method. Identified... The winding temperature can be calculated back from its temperature coefficient, and then compared with the rotor temperature. A motor thermal network model was established. These parameters were directly used for subsequent efficiency decoupling, avoiding errors caused by different operating conditions in offline measurements.

[0046] Specifically, the state vector estimated by the extended Kalman filter includes at least rotor speed, winding inductance, permanent magnet flux linkage and rotor temperature, with voltage, current, torque and mechanical speed as observations, and outputs the motor electromagnetic torque, various loss components and efficiency values ​​in real time.

[0047] Specifically, the thrust transient response indicators extracted in step S4 include the step response rise time t 90 Overshoot σ, settling time, and steady-state error; the temperature-thrust attenuation coefficient α is defined as: Where ΔT is the stator temperature rise. and These represent the thrust before and after the temperature rise, respectively.

[0048] In practical implementation, the segment containing the step command is extracted from the full mission cycle data collected in step three. After processing the thrust signal using a low-pass filter, the step initiation point is found using numerical differentiation. Using the steady-state initial value as a reference, the time required for the thrust to change from 10% to 90% is calculated as the rise time t_90; the peak thrust is then located, and the overshoot is calculated. in The steady-state thrust after the step jump is the average value within 2 seconds after the step jump. The adjustment time is the moment when the thrust first enters the range of F_ss±2% and no longer exceeds it.

[0049] The thermal attenuation coefficient α can be determined in a specialized test: first, a short-duration step test is performed at room temperature, and the cold thrust is recorded. and the stator temperature of the motor at that time Then, continue running the motor at its rated current until the stator temperature rises to a stable value. Then execute the same step command again and record the hot thrust. α is calculated according to the formula: Calculated, unit is ℃ -1 This coefficient can correct thrust measurements online and is used to evaluate thermal effects.

[0050] Specifically, in step S5, the vibration characteristic frequencies include at least the 1P and 2P frequencies corresponding to the fundamental frequency of the rotational speed, as well as the blade passage frequency (BPF); the vibration quality index includes the full-band vibration level in the 10~1000Hz frequency range and the amplitude of the aforementioned characteristic frequencies. The vibration velocity signal v(t) obtained from the laser vibrometer, after removing the trend term, undergoes a short-time Fourier transform at a sampling rate of 10kHz (Hanning window, window length 4096 points, step size 512 points, frequency resolution approximately 2.44Hz). The root mean square value of the velocity in the 10Hz to 1000Hz frequency band is calculated in each time frame and converted into a decibel vibration level. Reference speed Take 1×10 -6 m / s. Then, the real-time rotational frequency is obtained from the rotational speed signal: Along the order line in the time spectrum and Extract amplitude, where The number of propeller blades is used to obtain the vibration amplitude curves of these characteristic frequencies as a function of operating conditions through order slicing. The sideband modulation index is obtained by calculating the ratio of the energy within a specified bandwidth around the 1P peak to the energy of the 1P peak, and is used to evaluate the degree of modulation.

[0051] Specifically, in step S6, the task segment types include hovering, climbing, cruising, rapid maneuvering, and tilt transition. For each task segment, a subjective weight vector obtained by the analytic hierarchy process is preset, and then the objective weight is determined based on the current amount of data information using the entropy weight method. The final weight is obtained by multiplication. The calculation of the overall performance closeness adopts the grey relational analysis method.

[0052] In specific implementation, the combined weighting operation in step six is ​​as follows: For the hovering condition, the AHP judgment matrix can be calculated to give subjective weights W_sub=[0.2,0.4,0.3,0.1] corresponding to dynamic performance, energy efficiency, thermal stability, and vibration quality, respectively. The entropy weight method uses the distribution of each index in the current evaluation matrix to calculate the information entropy of the j-th index: in , , where n is the number of evaluation objects.

[0053] Objective weighting: Assuming that the calculation yields Given the value of [0.25, 0.35, 0.25, 0.15], the combined weight W = [0.2×0.25, 0.4×0.35, 0.3×0.25, 0.1×0.15] / (0.05 + 0.14 + 0.075 + 0.015) = [0.179, 0.500, 0.268, 0.053]. It is evident that the energy efficiency weight is highest when hovering.

[0054] Grey relational degree calculation: First, determine the reference sequence X_0=[1,1,1,1]. Calculate the grey relational coefficients of each indicator for each evaluation object: in ρ is set to 0.5. Then, the weighted average correlation degree is calculated. . This is the overall relevance score. Based on the relevance score, a rating and radar chart are generated.

[0055] Specifically, after step S4, there is also a health warning step: continuously monitor the residual between electromagnetic torque and mechanical torque and the sideband modulation of vibration characteristic frequency. When the residual exceeds the dynamic threshold or a specific fault characteristic appears, determine demagnetization, inter-turn short circuit or bearing wear fault, and associate it with the specific action in the task profile.

[0056] In a specific implementation, an improved embodiment of this method adds a health warning operation after step four and before step five. This involves continuously monitoring the electromagnetic torque output of the extended Kalman filter. Mechanical torque measured by torque sensor The difference forms a residual sequence. When the motor is functioning normally, the residual error mainly consists of model error and measurement noise. A dynamic threshold (e.g., 3 times the standard deviation) can be set by statistically analyzing its steady-state root mean square value. When the residual error exceeds the threshold or shows a trend of drift, combined with vibration signal characteristics: if there is a significant increase in vibration at twice the rotational frequency accompanied by a rotational frequency sideband, it is determined to be bearing wear; if the residual error is abnormal and frequency components related to the electrical frequency appear in the vibration, and the 1P vibration increases abnormally, it can be determined to be an inter-turn short circuit or demagnetization fault. All faults are associated with specific mission actions and time points, providing early warning information. If there is no fault, the resistance, flux linkage, and other parameters identified online by EKF are periodically updated to the UAV's digital twin model to maintain model accuracy and achieve condition-based maintenance.

[0057] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system, characterized in that, Includes the following steps: S1. Establish a typical flight mission profile library and convert it into dynamic throttle command sequences; S2. Build a multi-physical quantity synchronous test bench, and unify the acquisition clocks of thrust, torque, speed, voltage and current, multi-point temperature and vibration sensors to the same reference through a hardware synchronous triggering network based on satellite time synchronization and FPGA. S3. The real-time controller executes a dynamic command sequence with a fixed step size to drive the tested ESC-motor-propeller system and synchronously triggers all sensors to collect data, thereby obtaining a multi-parameter timing signal with a unified timestamp. S4. Adaptive filtering and interpolation alignment are performed on the multi-parameter time-series signal. The sensorless rotation speed is extracted using the current ripple. The electrical parameters and rotor temperature of the motor are estimated in conjunction with the extended Kalman filter. Then, the dynamic efficiency of the electronic speed controller, the electromagnetic efficiency and mechanical efficiency of the motor, and the aerodynamic efficiency of the propeller are decoupled and calculated. At the same time, the thrust transient response index and the temperature-thrust attenuation coefficient are extracted. S5. Extract the vibration level and characteristic frequency amplitude across the entire vibration band, and establish a four-dimensional evaluation matrix that includes dynamic performance, energy efficiency, thermal stability, and vibration quality; S6. Based on the task segment type to which the current dynamic instruction belongs, the weights of each dimension are adjusted in real time using a combined weighting method. The overall performance closeness is calculated by using grey relational analysis or the sorting method to approximate the ideal solution, and the performance level and multi-dimensional radar chart are output.

2. The method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to claim 1, characterized in that, In step S2, the hardware synchronization triggering network uses the second pulse and 10MHz clock output by the Beidou / GPS satellite receiving module as the system clock source, uses FPGA to generate multi-channel synchronization triggering pulses, and adds a timestamp containing absolute time and sampling sequence number to each sampling point. The sampling time deviation between different sensor channels is less than 50 nanoseconds.

3. The method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to claim 1, characterized in that, In step S4, the specific method for extracting sensorless rotational speed using current ripple is as follows: The DC bus current or three-phase AC current of the electronically controlled rectifier is collected, the commutation frequency is extracted from it, and the instantaneous rotational speed is obtained through a phase-locked loop or zero-crossing detection. Then Mechanical speed obtained from torque sensor Fusion, as an observation of the extended Kalman filter.

4. The method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to claim 3, characterized in that, The extended Kalman filter estimates the state vector, which includes at least rotor speed, winding inductance, permanent magnet flux linkage, and rotor temperature. Voltage, current, torque, and mechanical speed are used as observations, and the electromagnetic torque, loss components, and efficiency values ​​of the motor are output in real time.

5. The method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to claim 1, characterized in that, The thrust transient response parameters extracted in step S4 include the step response rise time t. 90 Overshoot σ, settling time, and steady-state error; the temperature-thrust attenuation coefficient α is defined as: Where ΔT is the stator temperature rise. and These represent the thrust before and after the temperature rise, respectively.

6. The method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to claim 1, characterized in that, In step S5, the vibration characteristic frequencies include at least 1P and 2P of the corresponding rotational speed fundamental frequency and the blade passing frequency BPF; the vibration quality index includes the full-band vibration level in the 10~1000Hz frequency range and the amplitude of the above characteristic frequencies.

7. The method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) power system according to claim 1, characterized in that, In step S6, the task segment types include hovering, climbing, cruising, rapid maneuvering and tilt transition. For each task segment, a subjective weight vector obtained by the analytic hierarchy process is preset, and then the objective weight is determined based on the current amount of data information using the entropy weight method. The final weight is obtained by multiplication synthesis. The grey relational analysis method was used to calculate the overall performance similarity.

8. A method for simultaneous testing and performance evaluation of multiple parameters of an unmanned aerial vehicle (UAV) propulsion system according to any one of claims 1 to 7, characterized in that, After step S4, a health warning step is also included: continuously monitoring the residual between electromagnetic torque and mechanical torque, as well as the sideband modulation of vibration characteristic frequency. When the residual exceeds the dynamic threshold or a specific fault characteristic appears, demagnetization, inter-turn short circuit, or bearing wear fault is determined and associated with the specific action in the task profile.

Images