A motor multi-parameter online monitoring and intelligent early warning method and system

Abstract

The application provides a motor multi-parameter online monitoring and intelligent early warning method and system. The method acquires original monitoring data, multiplexes an electronic speed regulator telemetry link, and collects control instructions, rotation speed, current, temperature, and vibration data by using an external lightweight sensor; time unification and preprocessing are performed on the original data, aligned data frames are generated by interpolation mapping and window statistics, and modal missing is marked; electrical characteristics, rotation speed characteristics, thermal characteristics, vibration characteristics, and control coupling residual characteristics are extracted to generate a characteristic vector; each modal confidence distribution is constructed, multi-modal fusion is performed after credibility correction, and a comprehensive risk value is generated; hierarchical early warning and protection control are implemented according to the comprehensive risk value and its persistence; baseline parameters are updated online in a healthy state, and event data is frozen, updated, and archived in an abnormal state. The application faces the strong coupling failure mechanism of a UAV motor, improves early fault identification capability, reduces false alarm rate, and supports flight safety closed-loop control.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) status monitoring and fault early warning, specifically to a method and system for online monitoring and intelligent early warning of multiple motor parameters. Background Technology

[0002] With the large-scale application of drones in industrial inspection, logistics transportation, and security monitoring, their flight safety and mission reliability are receiving increasing attention. As a critical component of the entire drone, the drone's power system, including the motor and electronic speed controller, is subjected to high dynamic loads, complex thermal fields, and mechanical vibration coupling during flight. Its health directly impacts flight safety. Therefore, online status sensing and early warning of anomalies for drone motors have become an important technological direction for improving the reliability of drone systems.

[0003] Existing technologies already include solutions for multi-parameter online monitoring and fault early warning. For example, one solution involves collecting temperature, current, vibration, acoustic, and environmental data, calculating anomalies, weighted fusion, and multi-level threshold early warning, and using a time synchronization protocol to construct a spatiotemporal dataset and global view. Another solution emphasizes the accumulation of long-term historical data for multiple parameters and big data analysis for motor health diagnosis. However, these existing solutions have significant limitations when directly applied to UAV motor monitoring scenarios and cannot meet practical needs.

[0004] Therefore, this application provides a method and system for online monitoring and intelligent early warning of multiple parameters of a motor to solve one of the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for online monitoring and intelligent early warning of multiple parameters of a motor, which can solve at least one of the technical problems mentioned above. The specific solution is as follows: According to a specific embodiment of this application, in a first aspect, this application provides a method for online monitoring and intelligent early warning of multiple parameters of a motor, comprising: The system acquires motor control command data output by the UAV flight controller, telemetry data returned by the electronic speed controller, and vibration and temperature data located in the motor mounting area. For the motor control command data, telemetry data, vibration data, and temperature data, data at different sampling rates are mapped into aligned data frames based on a unified timestamp, and modal validity markers are generated. Electrical features, rotational speed features, thermal features, vibration features, and control coupling features are extracted from the aligned data frames. Modal confidence distributions for healthy and fault states are constructed based on these features. The modal confidence distributions are then corrected for reliability according to modal integrity, noise level, and cross-modal consistency, and fused to obtain a comprehensive risk value. Based on the comprehensive risk value and the continuous over-threshold condition, early warning, alarm, or protection control results are output.

[0006] In one embodiment, the telemetry data returned by the electronic speed controller includes at least two of the following: motor speed, motor-side current, motor-side voltage, and electronic speed controller temperature; the vibration data is collected by an acceleration sensor installed on the motor mounting base, the end of the arm near the motor, or the motor housing; the temperature data is collected by a temperature sensor located on the motor housing, the heat conduction path of the motor mounting base, or the surface of the electronic speed controller; each motor has an independent data channel and motor identifier, so that control command data, telemetry data, vibration data, and temperature data are associated under the same motor identifier.

[0007] In one embodiment, mapping data at different sampling rates into aligned data frames based on a unified timestamp includes: constructing a target time sequence; interpolating and mapping low-sampling-rate data according to adjacent sampling points, and extracting window statistics and frequency domain values ​​for high-sampling-rate data according to a sliding time window corresponding to the target time; generating an invalidation flag for the corresponding mode when the time interval between the target time and the nearest valid sampling point exceeds the tolerance window; the aligned data frame includes at least the target time, rotational speed value, current value, voltage value, temperature value, vibration statistics, and validity flags for each mode.

[0008] In one embodiment, before generating the aligned data frame, the original data is further preprocessed; the preprocessing includes: outlier suppression and smoothing of current data and rotational speed data; bandpass filtering or narrowband analysis of vibration data around the fundamental frequency and harmonics of the motor rotation; low-pass filtering of temperature data; and setting rejection flags for abnormal data caused by communication packet loss, sensor saturation, or sampling interruption, wherein the rejection flags participate in subsequent modal reliability correction.

[0009] In one embodiment, the electrical characteristics include at least the mean current, root mean square current, rate of change of current, and rate of change of voltage; the speed characteristics include at least the mean speed, speed deviation, and rate of change of speed; the thermal characteristics include at least the temperature deviation of the motor housing, the rate of temperature rise, and the temperature difference between the electronic speed controller and the motor; the vibration characteristics include at least the root mean square vibration, peak factor, kurtosis, and narrowband harmonic energy around the fundamental frequency and harmonics of rotation; and the control coupling characteristics include at least a first residual between the control command and the actual speed and a second residual between the control command and the actual current.

[0010] In one implementation, the first residual and the second residual are calculated according to a health baseline model. The health baseline model takes control commands, motor-side voltage, and speed and current data under historical health conditions as inputs, and outputs the desired speed and desired current under the current control condition. The first residual is the difference between the actual speed and the desired speed, and the second residual is the difference between the actual current and the desired current. When the change in control commands is within a set range and both the first residual and the second residual increase simultaneously, it is determined to be an abnormal candidate state. When the change in control commands exceeds the set range, the corresponding time window is marked as a maneuver disturbance window and its contribution to the overall risk value is reduced.

[0011] In one embodiment, the step of constructing modal confidence distributions for healthy and faulty states based on various features, and then correcting and fusing the modal confidence distributions according to modal integrity, noise level, and cross-modal consistency to obtain a comprehensive risk value includes: generating modal confidence distributions for healthy states, mechanical imbalance faults, bearing deterioration faults, thermal anomalies, and electronic speed controller anomalies for electrical, thermal, vibration, and control coupling features, respectively; obtaining the confidence coefficient of each mode based on modal validity markers, signal quality evaluation results, and consistency evaluation results among modal features; correcting each modal confidence distribution using the confidence coefficients and then fusing them to obtain a comprehensive confidence distribution; determining the comprehensive risk value based on the confidence level corresponding to unhealthy states in the comprehensive confidence distribution, and stabilizing the comprehensive risk value for multiple consecutive target times.

[0012] In one embodiment, the step of outputting a warning, alarm, or protection control result based on a comprehensive risk value and a continuous threshold exceeding condition includes: setting a three-level risk threshold, including a first threshold, a second threshold, and a third threshold, within a range of 0 to 1; wherein the first threshold is lower than the second threshold, and the second threshold is lower than the third threshold; when the comprehensive risk value exceeds the first threshold for a continuous first duration, a warning result is output; when the comprehensive risk value exceeds the second threshold for a continuous second duration, an alarm result is output and a load reduction or return-to-base control suggestion is output; when the comprehensive risk value exceeds the third threshold for a continuous third duration, a protection control result is output; the protection control result includes at least one of limiting speed, limiting maneuvering amplitude, performing redundancy switching, or outputting a forced landing command; wherein the second duration is not lower than the first duration, and the third duration is not greater than the second duration.

[0013] In one implementation, when the overall risk value is lower than the first threshold, the current aligned data frame and the current feature are used as healthy samples, and the feature mean and feature fluctuation range in the healthy baseline model are updated using a forgetting factor; when the overall risk value reaches or exceeds the first threshold, the update of the healthy baseline model is frozen, and the aligned data frame, features, overall risk value evolution sequence, control command evolution sequence, and warning level before and after the anomaly are encapsulated into an event data packet, which is sent to a ground station, portable terminal, or cloud server for review and maintenance decision-making.

[0014] According to a specific embodiment of this application, in a second aspect, this application provides a multi-parameter online monitoring and intelligent early warning system for motors, comprising: The data access module is used to access motor control command data, electronic speed controller telemetry data, vibration data, and temperature data from the flight controller; the time unification module is used to map motor control command data, electronic speed controller telemetry data, vibration data, and temperature data with different sampling rates into aligned data frames and generate modal validity markers; the feature construction module is used to extract electrical features, speed features, thermal features, vibration features, and control coupling features; the risk fusion module is used to construct modal confidence distributions, perform confidence corrections, and output a comprehensive risk value; the early warning decision module is used to output early warning, alarm, or protection control results based on the comprehensive risk value and continuous over-threshold conditions; and the baseline update and archiving module is used to update the health baseline model and generate event data packets under abnormal conditions.

[0015] Compared with the prior art, the above-described solutions of this application have at least the following beneficial effects: This application provides a method for online monitoring and intelligent early warning of multiple motor parameters. By constructing a data acquisition, time unification, feature construction, risk fusion, early warning decision-making, and baseline update mechanism, it achieves specialized online monitoring of UAV motors and electronic speed controllers. Specifically, the data acquisition method, which reuses the telemetry link of the electronic speed controller and uses an external lightweight sensor, reduces system invasiveness and maintenance costs while ensuring multi-parameter coverage. The time unification module aligns and marks missing data from multi-rate heterogeneous data, laying a data foundation for subsequent fusion. The feature construction module extracts specialized features, including control coupling residuals, enabling the system to effectively distinguish between normal maneuvers and abnormal failures. The risk fusion module overcomes the shortcomings of traditional fixed-weighted methods in misjudging under multi-modal conflicts, missing data, or noise interference by fusing modal confidence distribution and credibility correction, significantly improving diagnostic robustness under complex flight conditions. The early warning decision-making module implements graded early warning and safety control by combining risk values ​​and state persistence, directly linking monitoring results to the flight safety closed loop. The baseline update module adaptively updates data in healthy states and freezes and archives event data in abnormal states, enhancing long-term environmental adaptability and preventing fault samples from contaminating the model. Overall, this application improves early fault identification capabilities, reduces false alarm rates, and achieves a closed-loop system from online perception to flight safety control without significantly increasing the hardware burden. Attached Figure Description

[0016] Figure 1 A flowchart of a method for online monitoring and intelligent early warning of multiple parameters of a motor is shown; Figure 2 A block diagram of a multi-parameter online monitoring and intelligent early warning system for motors according to an embodiment of this application is shown. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0019] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0020] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0021] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0022] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or system comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the product or system that includes said element.

[0023] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.

[0024] Currently, existing technologies include solutions for online monitoring of multiple parameters and fault early warning. However, these existing solutions have significant limitations when directly applied to UAV motor monitoring scenarios, specifically: The first approach to related technologies monitors the busbar, and its parameter system and abnormal mechanism revolve around the heating of conductive connection points, current harmonics and structural vibrations. It is difficult to reflect the failure characteristics of the UAV motor under the strongly coupled chain of "control command - motor speed - electrical load - mechanical vibration - thermal state" during flight. Moreover, its fixed weighted fusion method is prone to false alarms under the high-speed and variable operating conditions of the UAV.

[0025] The second approach to related technologies integrates multiple measurement units, but its structural design emphasizes pre-installation integration inside or on the surface of the motor. This is not conducive to the lightweighting, rapid replacement, and vibration tolerance of the UAV. Furthermore, it focuses on long-term health analysis and offline diagnosis, without addressing issues such as edge real-time decision-making, multi-rate heterogeneous data synchronization, false alarm suppression, and closed-loop control for flight safety.

[0026] Therefore, there is an urgent need to develop a specialized, multi-step, closed-loop, multi-parameter online monitoring and intelligent early warning technology solution for UAV motors.

[0027] In view of this, this application provides a method and system for online monitoring and intelligent early warning of multiple parameters of a motor to solve the above-mentioned technical problems.

[0028] The optional embodiments of this application are described in detail below with reference to the accompanying drawings.

[0029] The embodiments provided in this application are embodiments of a method for online monitoring and intelligent early warning of multiple parameters of a motor.

[0030] The following is combined with Figure 1 The embodiments of this application will be described in detail.

[0031] Figure 1 A flowchart of a method for online monitoring and intelligent early warning of multiple parameters of a motor is shown, such as... Figure 1 As shown, the procedure includes steps S101 to S106.

[0032] Step S101: Obtain motor control command data output by the UAV flight controller, telemetry data returned by the electronic speed controller, and vibration and temperature data set in the motor mounting area.

[0033] Step S102: For motor control command data, telemetry data, vibration data and temperature data, map each data with different sampling rates into aligned data frames based on a unified timestamp, and generate modal validity markers.

[0034] Step S103: Extract electrical features, rotational speed features, thermal features, vibration features, and control coupling features based on the aligned data frame.

[0035] Step S104: Construct modal confidence distributions for healthy and faulty states based on various features. Correct the confidence of the modal confidence distributions according to modal integrity, noise level, and cross-modal consistency, and then fuse them to obtain a comprehensive risk value.

[0036] Step S105: Output early warning, alarm or protection control results based on the comprehensive risk value and the continuous over-threshold condition.

[0037] Step S106: When the comprehensive risk value is lower than the first threshold, the current aligned data frame and the current feature are used as healthy samples, and the feature mean and feature fluctuation range in the healthy baseline model are updated using the forgetting factor; when the comprehensive risk value reaches or exceeds the first threshold, the update of the healthy baseline model is frozen, and the aligned data frame, features, comprehensive risk value evolution sequence, control command evolution sequence and warning level before and after the anomaly are encapsulated into an event data packet and sent to the ground station, portable terminal or cloud server for review and maintenance decision.

[0038] As a specific embodiment, step S101 is used to construct the original monitoring data acquisition mechanism. The input data for this step includes: motor control command data output by the UAV flight controller, telemetry data returned by the electronic speed controller, vibration sampling data output by an external vibration sensor, temperature sampling data output by an external temperature sensor, and acoustic sampling data output by an optional acoustic sensor. The processing procedure for this step is as follows: At least two of the following are prioritized for acquisition: motor speed, motor-side current, motor-side voltage, and electronic speed controller temperature, which are obtained from the existing telemetry link of the electronic speed controller. Based on this, a vibration sensor is installed near the motor mounting base on the arm to obtain mechanical responses that characterize bearing deterioration, dynamic balance degradation, and structural resonance. Specifically, vibration data is collected by an acceleration sensor installed on the motor mounting base, the arm end near the motor, or the motor housing. A temperature sensor is installed on the heat conduction path of the motor housing or motor mounting base to compensate for the deficiency that the electronic speed controller temperature cannot adequately reflect the thermal state of the motor body. Specifically, temperature data is collected by a temperature sensor installed on the heat conduction path of the motor housing, the motor mounting base, or the surface of the electronic speed controller. When noise conditions permit, acoustic sensors should be configured to acquire sound characteristics of fault howling, scraping, and localized impacts.

[0039] Unlike the second approach in related technologies, which involves "pre-installing each monitoring unit inside the motor or mounting it on the motor," this application adopts a structure of "telemetry multiplexing + external lightweight sensing," decoupling the monitoring system from the UAV motor body. This avoids the problem of replacing the entire monitoring device when the motor is replaced, while also reducing installation complexity. Each motor has an independent data channel and motor identifier, linking control command data, telemetry data, vibration data, and temperature data under the same motor identifier.

[0040] For example, the output of step S101 is: a set of original monitoring data with uniform parameter definitions, denoted as D0={u(t), r(t), i(t), v(t), T(t), a(t), s(t)}. Here, u(t) represents control command data, r(t) represents rotational speed data, i(t) represents current data, v(t) represents voltage data, T(t) represents temperature data, a(t) represents vibration data, and s(t) represents acoustic data. If a mode does not exist, it is treated as an empty mode, rather than requiring all modes to exist simultaneously.

[0041] For example, on one of the powered arms of a quadcopter drone, the flight controller outputs throttle control commands to the electronic speed controller (ESC). The ESC periodically returns the real-time speed, output current, and ESC temperature of the motor. A miniature accelerometer is attached to the motor mount of the powered arm, and a thermistor is attached to the outer wall of the motor housing. Thus, the powered arm can generate a raw monitoring data set containing five categories of raw monitoring data: control commands, speed, current, temperature, and vibration. This implementation does not require disassembling the internal structure of the motor, and therefore is more suitable for on-site maintenance and replacement of drones compared to the built-in structure of the second approach in related technologies.

[0042] In some embodiments, step S102 is used to perform time unification and preprocessing on the raw monitoring data to generate aligned data frames.

[0043] For example, the input data for step S102 is the raw monitoring data set D0 output from step S101.

[0044] The processing procedure in step S102 is as follows: Since the sampling rates of speed, current, temperature, vibration, and acoustic data are usually different, direct fusion judgment cannot be made. This application assigns timestamps to each type of raw monitoring data and constructs a target time sequence. For low sampling rate data, such as temperature data and some electronic speed controller telemetry data, an interpolation method is used to map to the target time. For high sampling rate data, such as vibration data and acoustic data, direct point-to-point alignment is not performed; instead, windowing processing is performed within the corresponding time window, and then the statistics or frequency domain quantities corresponding to the target time are output. If a certain mode is missing at a certain time, no forced filling is performed; instead, a missing marker is generated to reduce the weight of that mode during subsequent fusion. Its linear interpolation formula can be expressed as: ; in, and These are the sampled values ​​of the mode to be interpolated at two adjacent sampling times, respectively. and These are the corresponding timestamps. For the target time The interpolation results are used. If the time interval between the target time and the nearest valid sampling point exceeds the tolerance window, an invalid flag is generated for that mode instead of outputting the interpolated value. The aligned data frame includes at least the target time, rotational speed, current, voltage, temperature, vibration statistics, and validity flags for each mode.

[0045] Before generating aligned data frames, this step also preprocesses each mode. Preprocessing includes: outlier suppression and smoothing of current and rotational speed data; bandpass filtering or narrowband analysis of vibration data around the motor's fundamental frequency and its harmonics; low-pass filtering of temperature data; and setting rejection flags for abnormal data caused by communication packet loss, sensor saturation, or sampling interruptions. These rejection flags participate in subsequent mode reliability correction. Here, filtering parameters are set based on the structural characteristic frequencies of the monitored motor, the propeller passing frequencies of the UAV, and the resonance range of the arm structure. If specific values ​​need to be set, they are determined based on the results of pre-experimental spectrum analysis. For example, if the first-order rotational frequency of a UAV motor under typical cruise conditions is concentrated in several known intervals, the center frequency band of the vibration bandpass filter can be set around these frequencies and their harmonics. Since the frequency distribution varies significantly under different motor pole pairs, propeller sizes, and control strategies, this application does not provide uniform fixed values.

[0046] For example, the output of step S102 is: an aligned data frame set. Each of the aligned data frames All include valid telemetry data, window statistics, and modal validity markers at the same target time.

[0047] For example, at a target time tk, the electronic speed controller uploads speed, current, and temperature data once, while vibration data has already been sampled extensively within a short time window before and after tk. In this step, the speed, current, and temperature are directly mapped to that frame. Simultaneously, the vibration time-domain statistics and narrowband energy are extracted from this time window as the vibration components corresponding to that frame, and the presence of acoustic modes is recorded. Thus, subsequent steps deal not with asynchronous raw sampling streams, but with aligned data frames with a uniform structure.

[0048] As a feasible embodiment, step S103 is used to perform multi-parameter feature extraction on the aligned data frame to generate a feature vector.

[0049] For example, the input data for step S103 is the set of aligned data frames F output in step S102 and the set of healthy baseline parameters B maintained in step S106.

[0050] For example, the processing procedure in step S103 is as follows: This application does not follow the first solution of the related technology for object characteristics such as surface temperature anomaly of the connection point, axial temperature gradient anomaly, and three-phase imbalance of the busbar, but instead constructs a feature system for UAV motors. This feature system includes electrical characteristics, rotational speed characteristics, thermal characteristics, vibration characteristics, control coupling characteristics, and optional acoustic characteristics.

[0051] The electrical characteristics include at least the mean current, root mean square current, rate of change of current, and rate of change of voltage, denoted as Ime, Irms, and dI / dt, respectively. This is because drone motor failures during flight more often manifest as abnormal current spikes, load fluctuations, and transient pulses, rather than just harmonics and three-phase imbalances as seen in industrial power supply scenarios.

[0052] For example, the speed characteristics include at least the mean speed, speed deviation, and speed change rate, denoted as Rme, ΔR, and dR / dt. The deviation between the speed and the control command is the key to distinguishing between "normal maneuvering" and "failure".

[0053] For example, thermal characteristics include at least the temperature deviation of the motor casing, the rate of temperature rise, and the temperature difference between the electronic speed controller and the motor, denoted as ΔT, dT / dt, and ΔTes. Since the thermal state of the UAV motor is significantly affected by airflow cooling and load changes, the absolute temperature value alone cannot reliably determine the cause of failure.

[0054] For example, vibration characteristics include at least the root mean square (RMS) of vibration, peak factor, kurtosis, and narrowband harmonic energy around the fundamental and harmonic frequencies. The formula for the RMS of vibration is: ; in, These are the vibration sampling values ​​within the time window.

[0055] The kurtosis formula is: ; in, This represents the average vibration sampling value within this time window.

[0056] For example, to reduce embedded implementation overhead, this application prioritizes narrowband harmonic energy over full-frequency FFT energy, and its Goertzel calculation formula can be written as: ; ; in, This represents the state variable in the Goertzel recursion process. The angular frequency corresponding to the target frequency is represented by... The calculation yielded, where The frequency of the target to be detected. The sampling frequency is used because the fault characteristics of UAV motors are often concentrated near the fundamental frequency and its harmonics. If the first solution of related technologies is used to process the FFT energy distribution of the entire frequency band from 0.1Hz to 5kHz, it will increase the onboard computing burden and is not conducive to low-power deployment.

[0057] For example, the control coupling feature is a key feature that distinguishes this application from related technologies, and it includes at least a first residual between the control command and the actual rotational speed and a second residual between the control command and the actual current. It can be defined as follows: ; ; in, Indicates the first Residual rotational speed at time step Indicates the first Current residual at time , Represents a discrete-time index. Indicates the first The actual motor speed at time k, and i(k) represents the actual motor current at time k. Indicates the first Time control commands, This represents the motor-side voltage at time k. This indicates the control command u(k) and voltage. Predicted expected rotational speed under the given conditions Indicates in control commands Actual speed and voltage The expected current prediction under the given conditions. This expected value can be given by the healthy baseline model.

[0058] For example, the health baseline model takes control commands, motor-side voltage, and historical speed and current data under healthy conditions as input, and outputs the desired speed and current under the current control condition. Its core idea is: if the current increases synchronously with the control command while the speed is normal, it is more likely a normal maneuver. If the control command remains basically unchanged but the current increases abnormally and the speed decreases, it is more likely an abnormal mechanical load or an electronic speed controller malfunction. When the control command change is within the set range and the first residual and the second residual increase simultaneously, it is identified as an abnormal candidate state. When the control command change exceeds the set range, the corresponding time window is marked as a maneuver disturbance window, and its contribution to the overall risk value is reduced.

[0059] For example, if an acoustic mode is configured, short-time energy, characteristic frequency band energy, or cepstral features can be extracted. However, this mode is preferentially used as an auxiliary mode and not as the primary criterion because the wind noise in flight is highly uncertain.

[0060] For example, the output of step S103 is a multi-parameter feature vector: .

[0061] For example, if a motor exhibits minimal changes in control commands during hovering, but shows a slight decrease in speed, a significant increase in current, an increase in the root mean square of vibration, and enhanced narrowband frequency harmonic energy, while the casing temperature has not yet risen significantly, this step will output a feature vector indicating "large control coupling residual + significant vibration anomaly + no thermal anomaly yet formed." This feature vector can characterize early bearing degradation or blade dynamic balance problems, a feature not defined for the UAV object in the first approach of the related technology.

[0062] In some embodiments, step S104 is used to construct a modal confidence distribution based on the feature vector and perform confidence correction fusion to generate a comprehensive risk value.

[0063] For example, the input data for step S104 is the feature vector zk output in step S103, the modality validity label output in step S102, and the set of health baseline parameters B output in step S106.

[0064] For example, the processing procedure in step S104 is as follows: Unlike the first approach in related technologies, which directly uses weighted average fusion of temperature, vibration, acoustic, and current feature scores, this application first constructs a modal confidence distribution for each mode, and then performs confidence-corrected fusion to solve the problems of "temporary distortion of a certain mode, missing of a certain mode, and conflict of evidence between different modes" during UAV flight. The first approach in related technologies explicitly uses a weighted average fusion algorithm to calculate the confidence score of local faults and introduces an environmental correction factor η and weight coefficients for each mode. The drawback of this method is that although the weights are adjustable, their essence is still linear superposition, which is not robust enough for highly conflicting evidence.

[0065] This application defines a set of fault assumptions. .in, Indicates health status. This indicates a mechanical imbalance-related fault. This indicates a bearing deterioration-related fault. This indicates a thermal or winding abnormality. This indicates an abnormality in the electronic speed controller or power supply link. Modal confidence distributions are generated for corresponding modes based on electrical, thermal, vibration, and control coupling characteristics, representing health status, mechanical imbalance faults, bearing deterioration faults, thermal anomaly faults, and electronic speed controller anomaly faults, respectively. If a mode is missing in the current frame, it is not forcibly assigned a value; instead, more quality is allocated to the uncertain set. Then, based on the modal validity label, signal quality evaluation results, and consistency evaluation results among the modal characteristics, the confidence coefficient of each mode is obtained. Its value is determined by the normalized result output by the quality evaluation function. However, due to differences in communication quality, airframe structural noise, and sensor installation conditions on different UAV platforms, fixing the confidence coefficient may weaken its universality.

[0066] Subsequently, the confidence distribution for each modality is adjusted for confidence level: ; ; in, Indicates the first Each mode in state The confidence level after confidence adjustment is as follows. Indicates the first The confidence coefficient of each modality Indicates the first Each mode in state The original confidence level is as follows. Represents any specific state in the set of fault states. Represents the set of states of the entire set. Indicates the first The confidence level of the residual uncertainty of each mode in the whole set. It represents the sum of the confidence levels of each specific state other than the overall set state.

[0067] Then, multimodal fusion is performed sequentially to generate a comprehensive confidence distribution. The overall risk value is defined as: ; Or defined as: ; in, Indicates the first The overall risk value at any given moment. Indicates the first Confidence level of health status after real-time integration. Indicates the first After the moment merges Confidence level of fault states, Indicates health status. Indicates the first Fault-like states, For different fault categories, the risk severity coefficient represents the first... Risk weight coefficients corresponding to different failure states Represents a discrete-time index. Indicates modal index, This indicates the fault category index. If a specific value needs to be set, it will be based on the drone safety strategy analysis. For example, the risk of single-propeller failure caused by electronic speed controller synchronization loss is generally higher than that of simple temperature rise, therefore... The corresponding severity coefficient should be higher than This value is given by the conclusions of flight risk analysis or failure mode and effects analysis.

[0068] For example, to further reduce false alarms, this step performs time stabilization on the composite risk value. A sliding window of length W is defined, and the median of the composite risk values ​​for the most recent W frames is taken to obtain a stable composite risk value. The median, rather than the mean, is used because sudden control disturbances occur during UAV flight, and the median is more robust to isolated pulses. The value of W is determined based on the target alarm delay limit and noise level: if the system requires alarms to be completed in a short time, W should not be too large. If the system is more concerned with suppressing short-pulse false alarms, W can be increased appropriately.

[0069] For example, the output of step S104 is the stable composite risk value. and the corresponding main fault category judgment results .

[0070] For example, within a certain flight time window, electrical characteristics indicate an abnormal current surge, but the control coupling residual is small, suggesting that it was caused by a transient flight control maneuver. Simultaneously, if there is no abnormal increase in vibration and thermal modes, the electrical mode conflicts significantly with other modes, its reliability will be lowered, and the overall risk value will not increase significantly. Conversely, if the control command changes little but vibration kurtosis, narrowband harmonic energy, and current anomalies increase simultaneously, the multimodal evidence tends to be consistent, and the overall risk value increases rapidly. This approach is more adaptable to the dynamic flight conditions of UAVs than the fixed-weighted average method of the first scheme in related technologies.

[0071] As a specific embodiment, step S105 is used to perform graded early warning and protection control based on the comprehensive risk value and the persistence of the status, and generate early warning results.

[0072] For example, the input data for step S105 is the stable comprehensive risk value output in step S104. and the results of the main fault category judgment .

[0073] For example, the processing procedure in step S105 is as follows: This application adopts a graded early warning strategy, but does not mechanically follow the three-level fixed threshold scheme with typical values ​​of 0.7, 1.5, and 3.0 in the first scheme of the related technology. The reason is that the local fault confidence score in the first scheme of the related technology is generated based on its own anomaly calculation and weighted fusion framework, and its scoring scale may not be consistent with the stable comprehensive risk value of this application. If it is directly copied, it will cause incompatibility of the scoring space. Therefore, this application sets a three-level risk threshold, including a first threshold, a second threshold, and a third threshold, in the range of 0 to 1. Among them, the first threshold is lower than the second threshold, and the second threshold is lower than the third threshold.

[0074] For example, the decision logic for step S105 is as follows: When At that time, it is determined to be a healthy state. When Furthermore, if the duration exceeds the first threshold for the first consecutive period, an early warning result is output, indicating the presence of an early anomaly. When And if the second consecutive duration exceeds the second threshold, an alarm result is output, and a load reduction suggestion or return-to-base suggestion is sent to the flight control system. Furthermore, when the third consecutive duration exceeds the third threshold, a protection result is output, and a safety control strategy is triggered, such as limiting engine speed, restricting maneuvering amplitude, performing redundancy switching, or outputting a forced landing command. The duration frames N1, N2, and N3 are introduced as duration periods to avoid false triggering caused by single-frame pulse values. Their specific values ​​are determined based on the system sampling period, the maximum allowable alarm delay, and the fault evolution speed. The second duration period is not less than the first duration period, and the third duration period is not greater than the second duration period.

[0075] For example, the output of step S105 is the warning result. It includes the warning level, the main fault category, the duration, the recommended action, and the timestamp.

[0076] For example, during a patrol flight, if the overall risk value of a certain power arm motor exceeds the first-level threshold for multiple consecutive target moments, but does not reach the second-level threshold, the system outputs "Level 1 Warning: Suspected early signs of mechanical imbalance; it is recommended to check the propellers and mounting base after the mission ends." If the overall risk value further increases and continuously exceeds the second-level threshold, it outputs "Level 2 Alarm: Suspected worsening of mechanical imbalance; it is recommended to immediately reduce the load and return to base." This type of warning output is directly related to the closed-loop control of the UAV mission, which is significantly different from the second approach of related technologies, which is biased towards long-term operation and maintenance analysis structures on portable devices and cloud servers.

[0077] In some embodiments, step S106 is used to perform online baseline updates on low-risk operational data and to perform event archiving and offline model correction on high-risk event data.

[0078] For example, the input data for step S106 is the warning result output in step S105. The stable comprehensive risk value output in step S104 The feature vector zk output in step S103 and the set of aligned data frames F output in step S102.

[0079] For example, the processing procedure in step S106 is as follows: When the comprehensive risk value is lower than the first threshold, the current aligned data frame and the current feature are used as healthy samples, and the feature mean and feature fluctuation range in the healthy baseline model are updated using a forgetting factor. When the comprehensive risk value reaches or exceeds the first threshold, the healthy baseline model update is frozen, and the aligned data frame, features, comprehensive risk value evolution sequence, control command evolution sequence, and warning level before and after the anomaly are encapsulated into an event data packet, which is sent to the ground station, portable terminal, or cloud server for review and maintenance decision-making. Online baseline updates can use an exponential sliding update method. ; ; in, This represents the updated baseline of the feature mean. This represents the baseline of the feature mean before the update. This represents the updated baseline of characteristic variance. This represents the baseline of the characteristic variance before the update. Indicates the forgetting factor, This represents the feature value of the current input. This represents the squared deviation between the current feature value and the updated mean. The settings are determined based on the rate of change in operating conditions: if the drone's mission conditions remain stable over a long period, then... Set it to a larger value to improve baseline stability. If the drone mission conditions change rapidly, then... Reduce it appropriately to improve the speed of baseline adaptation.

[0080] For example, when the system enters a warning, alarm, or protection state, the aligned data frames before and after the current time window, feature vectors, the evolution process of the comprehensive risk value, and the evolution process of the control commands will be compiled into an event data packet for archiving. This event data packet can be sent to a ground station or the cloud for fault review, maintenance decisions, and subsequent model parameter correction. This approach incorporates the advantages of the second solution in related technologies, namely "interaction between portable devices and cloud servers, big data analysis of long-term historical operating data, formulation of maintenance plans, and analysis of design defects," but this application positions it as a gain link after real-time closed-loop processing, rather than an essential path for real-time alarms.

[0081] For example, the output of step S106 is the updated set of health baseline parameters Bnew, and the event data packet Event.

[0082] For example, during multiple missions, the current fluctuation range and root mean square vibration of a certain type of UAV motor under healthy conditions may slowly drift due to slight differences in seasonal temperature and propeller performance. If the system consistently uses the factory-fixed baseline, false alarms will increase over time. After adopting the online baseline update in this step, the system can gradually correct the normal fluctuation range of current and vibration under healthy conditions. When a significant anomaly occurs during a flight, the system freezes the baseline update and packages and stores the data before and after the event for maintenance personnel to confirm whether it is propeller damage, bearing deterioration, or electronic speed controller failure. This "adaptive in healthy conditions, frozen and logged in abnormal conditions" approach represents a significant improvement to related technologies.

[0083] In some embodiments, steps S101 to S106 in the online multi-parameter monitoring and intelligent early warning method for motors can be summarized into a data acquisition step, a time unification step, a feature construction step, a risk fusion step, an early warning decision step, and a baseline update step based on their execution content. Specifically, the data acquisition step acquires motor control command data, electronic speed controller telemetry data, vibration sampling data, temperature sampling data, and optional acoustic sampling data, and outputs a raw monitoring data set. The time unification step performs timestamp unification, interpolation mapping, window statistics, and modal missing markers on the raw monitoring data set, and outputs an aligned data frame. The feature construction step extracts electrical features, speed features, thermal features, vibration features, control coupling features, and optional acoustic features based on the aligned data frame, and outputs a feature vector. The risk fusion step constructs modal confidence distributions for each modality for healthy and fault states based on the feature vectors, corrects the confidence of each modal confidence distribution, and generates a comprehensive risk value through multi-modal fusion. The early warning decision step determines the early warning level based on the comprehensive risk value and its persistence, and outputs action suggestions corresponding to the fault category. The baseline update step is used to update baseline parameters in a healthy state and freeze the baseline and generate event data packets in an abnormal state.

[0084] As a feasible implementation, the time unification step employs interpolation for low-frequency modes and sliding time window feature extraction for high-frequency modes, uniformly mapping all modes onto the target time sequence. Low-frequency modes include at least temperature data and some electronic governor telemetry data, while high-frequency modes include at least vibration and acoustic data. The time unification step also outputs modal validity markers for credibility correction in the risk fusion step.

[0085] As a specific embodiment, the control coupling features extracted in the feature construction step include at least the residual deviation between the control command and the actual rotational speed, and the residual deviation between the control command and the actual current. This residual deviation is composed of the difference between the expected value given by the healthy baseline model and the current observation value, and is used to distinguish between parameter changes caused by normal maneuvering and parameter changes caused by faults.

[0086] For example, in some embodiments, the risk fusion step does not employ a simple weighted average method. Instead, it first generates modal confidence distributions for electrical, thermal, vibrational, and acoustic modes separately. Then, it corrects the confidence distributions for each mode based on modal integrity, noise level, and modal consistency, and finally generates a comprehensive risk value. Thus, even if a mode is temporarily missing or subjected to strong noise interference, it will not directly cause the comprehensive risk value to be distorted.

[0087] As a feasible implementation, the early warning decision-making process employs a three-level early warning mechanism, including a level-one early warning, a level-two alarm, and a level-three protection. The level-one early warning is used to indicate early anomalies and suggest maintenance after the mission ends. The level-two alarm is used to indicate escalating anomalies and suggest immediately reducing the load or returning to base. The level-three protection is used to indicate high-risk conditions and output protection control commands. The early warning decision-making process also uses the number of consecutive frames exceeding the threshold of the comprehensive risk value as one of the triggering conditions to suppress false triggers caused by isolated anomalies.

[0088] As one specific implementation, the baseline update step updates the mean baseline and fluctuation baseline only when the overall risk value is below the first-level warning threshold, and freezes the update once the overall risk value reaches the first-level warning threshold. This setting is used to prevent faulty samples from entering the normal baseline, thereby maintaining the accuracy of subsequent anomaly identification.

[0089] This application discloses a method for online monitoring and intelligent early warning of multiple parameters of a motor. This method is aimed at the motor and its associated electronic speed controller in the power system of a UAV, performing online status perception, anomaly identification, risk assessment, and early warning output. Compared with the first solution of related technologies, which discloses a "multi-parameter online monitoring and fault early warning system based on intelligent busbars", its core technical path is still based on the acquisition of temperature, current, vibration, acoustic and environmental data, anomaly calculation, weighted fusion, and multi-level threshold early warning, and constructs a spatiotemporal dataset and global view through a time synchronization protocol. However, the monitoring object of this type of solution is the busbar, and its anomaly mechanism revolves around the heating of conductive connection points, current harmonics, structural vibration, and acoustic emission. It lacks targeted modeling of the strongly coupled chain of "control command - motor speed - electrical load - mechanical vibration - thermal state" in the operation of UAV motors, so it is difficult to directly apply to the online monitoring scenario of UAV motors. On the other hand, while the second approach to related technologies has disclosed built-in speed measurement units, temperature measurement units, and electrical parameter measurement units, as well as integrated intelligent operation monitoring and early warning devices, integrated intelligent bearing monitoring and early warning devices, and integrated intelligent vibration monitoring and early warning devices, and introduces portable devices and cloud servers into the health diagnosis and early warning system, it emphasizes the factory-integrated structure of the motor, the accumulation of long-term historical data of multiple parameters, and big data analysis. It does not specifically address lightweight deployment of UAVs, edge real-time decision-making, multi-rate heterogeneous data synchronization, false alarm suppression, and closed-loop control for flight safety. Therefore, this application, in its design phase, does not simply piece together existing technologies, but rather, based on identifying the aforementioned shortcomings, constructs a specialized, multi-step closed-loop, multi-parameter online monitoring and intelligent early warning technology solution for UAV motors.

[0090] The significant technical differences between this embodiment and related technologies include at least the following: First, the monitoring object is generalized from busbars or industrial motors to UAV motors and their associated electronic speed controllers. The monitoring logic revolves around the coupling chain of "control command—speed—current—vibration—thermal state," rather than around the temperature rise at the busbar connection point and branch current harmonics. This limitation on the object and the mechanism constitute a significant difference. Second, the data acquisition architecture adopts a combination of "electronic speed controller telemetry multiplexing + external lightweight sensors," rather than a fixed integrated acquisition structure built into the motor, thus adapting to the lightweight and rapid maintenance scenarios of UAVs. Third, at the data processing level, a unified time frame and missing data marker mechanism for multi-rate heterogeneous data are introduced, which is not specifically tailored to the UAV scenario in related technologies. Fourth, at the feature level, control coupling residuals and narrowband harmonic energy are introduced, enabling the system to distinguish between normal maneuvers and abnormal failures. Fifth, at the fusion level, a modal confidence distribution, confidence correction, and stable comprehensive risk value generation mechanism are used to replace the fixed weighted average method in the first scheme of related technologies. Sixth, at the strategic level, the online early warning closed loop and the offline event analysis closed loop are decoupled, which not only absorbs the long-term operation and maintenance advantages of the second solution of related technologies, but also ensures that real-time response during flight does not depend on the cloud.

[0091] The beneficial effects of each step in this embodiment are as follows: Step S101 improves parameter coverage and reduces intrusion into the motor structure by combining telemetry multiplexing and external sensing. Step S102 improves the fusionability of multi-source data through time unification, interpolation, and missing data labeling. Step S103 improves fault sensitivity and reduces feature redundancy by constructing dedicated features for UAV motor mechanisms. Step S104 improves anti-collision, anti-missing, and anti-noise capabilities through credibility-corrected fusion. Step S105 transforms the identification results into executable flight safety actions through hierarchical early warning and continuous decision-making. Step S106 balances long-term adaptability and model purity through adaptive updates in healthy states and freeze-keeping in abnormal states. The overall technical solution is beneficial in that it enables continuous online monitoring of motor anomalies during UAV flight, improves early fault identification capabilities without significantly increasing hardware burden, reduces false alarms caused by normal maneuvers, and directly correlates identification results with flight safety control and subsequent maintenance decisions, thereby achieving the target technical effect of "online monitoring and intelligent early warning of multiple parameters of UAV motors".

[0092] This application also provides system embodiments that follow the above embodiments, for implementing the method steps of the above embodiments. The interpretation of the same names is the same as that of the above embodiments, and they have the same technical effects as those of the above embodiments, so they will not be repeated here.

[0093] like Figure 2As shown, this application provides a multi-parameter online monitoring and intelligent early warning system 200 for motors, comprising: The data access module 201 is used to access the motor control command data of the flight controller, the telemetry data of the electronic speed controller, the vibration data, and the temperature data.

[0094] The time unification module 202 is used to map motor control command data, electronic speed controller telemetry data, vibration data and temperature data with different sampling rates into aligned data frames and generate modal validity markers.

[0095] Feature construction module 203 is used to extract electrical features, rotational speed features, thermal features, vibration features, and control coupling features.

[0096] Risk fusion module 204 is used to construct modal confidence distributions, perform confidence corrections, and output a comprehensive risk value.

[0097] The early warning decision module 205 is used to output early warning, alarm or protection control results based on the comprehensive risk value and the continuous over-threshold condition.

[0098] The baseline update and archiving module 206 is used to update the health baseline model and generate event data packets in abnormal states.

[0099] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0100] Although the operations are described in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all of the operations shown to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0101] The methods and systems of this application can be implemented using standard programming techniques, utilizing rule-based logic or other logic to implement various method steps. It should also be noted that the terms "system" and "module" as used herein and in the claims are intended to include implementations using one or more lines of software code and / or hardware implementations and / or devices for receiving input.

[0102] Any step, operation, or procedure described herein may be performed or implemented using one or more hardware or software modules, either alone or in combination with other devices. In one embodiment, the software module is implemented using a computer program product comprising a computer-readable medium containing computer program code, which is executable by a computer processor to perform any or all of the described steps, operations, or procedures.

[0103] The foregoing description of implementations of this application has been provided for illustrative and descriptive purposes. The foregoing description is not exhaustive and is not intended to limit this application to the exact forms disclosed. Various modifications and variations may exist in accordance with the foregoing teachings, or may arise from practice of this application. These embodiments were chosen and described to illustrate the principles of this application and its practical application, enabling those skilled in the art to utilize this application in various implementations and modifications to suit the specific purpose of the concept.

[0104] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0105] It can be further understood that, unless otherwise specified, "connection" includes both direct connections where no other components exist between the two parties and indirect connections where other components exist between them.

[0106] It is further understood that although the operations are described in a specific order in the accompanying drawings in the embodiments of this application, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all the operations shown to be performed to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0107] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the field of this application that are not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0108] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

[0109] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for online monitoring and intelligent early warning of multiple parameters of a motor, characterized in that, include: Acquire motor control command data output by the UAV flight controller, telemetry data returned by the electronic speed controller, and vibration and temperature data set in the motor mounting area; For the motor control command data, the telemetry data, the vibration data, and the temperature data, each of the data at different sampling rates is mapped into an aligned data frame based on a unified timestamp, and a modal validity marker is generated; Based on the aligned data frame, electrical features, rotational speed features, thermal features, vibration features, and control coupling features are extracted. Based on the features described in each category, modal confidence distributions for healthy and faulty states are constructed respectively. The confidence distributions are then corrected for confidence based on modal integrity, noise level, and cross-modal consistency, and fused to obtain a comprehensive risk value. Based on the comprehensive risk value and the continuous over-threshold condition, output early warning, alarm or protection control results.

2. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 1, characterized in that, The telemetry data returned by the electronic speed controller includes at least two of the following: motor speed, motor-side current, motor-side voltage, and electronic speed controller temperature. The vibration data is collected by an acceleration sensor installed on the motor mounting base, the end of the arm near the motor, or the motor housing. The temperature data is collected by temperature sensors located on the motor housing, the heat conduction path of the motor mounting base, or the surface of the electronic speed controller. Each motor has an independent data channel and motor identifier, which associates control command data, telemetry data, vibration data and temperature data under the same motor identifier.

3. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 2, characterized in that, The mapping of data at different sampling rates to aligned data frames based on a unified timestamp includes: Construct the target time series; For low sampling rate data, interpolation mapping is performed based on adjacent sampling points; for high sampling rate data, window statistics and frequency domain parameters are extracted based on the sliding time window corresponding to the target time. When the time interval between the target time and the nearest valid sampling point exceeds the tolerance window, an invalid flag is generated for the corresponding mode. The aligned data frame includes at least the target time, rotational speed, current, voltage, temperature, vibration statistics, and validity markers for each mode.

4. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 3, characterized in that, Before generating the aligned data frame, the process also includes preprocessing the raw data; The preprocessing includes: Outlier suppression and smoothing were performed on current and speed data; Bandpass filtering or narrowband analysis of the vibration data around the fundamental frequency and its harmonics of the motor rotation; Apply low-pass filtering to the temperature data; Abnormal data caused by communication packet loss, sensor saturation, or sampling interruption is marked for rejection, and the rejection mark participates in subsequent modal reliability correction.

5. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 4, characterized in that, The electrical characteristics include at least the mean current, root mean square current, rate of change of current, and rate of change of voltage. The speed characteristics include at least the mean speed, speed deviation, and speed change rate; The thermal characteristics include at least the temperature deviation of the motor housing, the rate of temperature rise, and the temperature difference between the electronic speed controller and the motor. The vibration characteristics include at least the root mean square of the vibration, the peak factor, the kurtosis, and the narrowband harmonic energy around the fundamental frequency and its harmonics. The control coupling feature includes at least a first residual between the control command and the actual rotational speed and a second residual between the control command and the actual current.

6. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 5, characterized in that, The first and second residuals were calculated according to the healthy baseline model; The health baseline model takes control commands, motor-side voltage, and historical speed and current data under health conditions as inputs, and outputs the desired speed and desired current under the current control conditions. The first residual is the difference between the actual rotational speed and the desired rotational speed, and the second residual is the difference between the actual current and the desired current; When the change in the control command is within the set range and the first residual and the second residual increase simultaneously, it is determined to be an abnormal candidate state; When the change in control commands exceeds the set range, the corresponding time window is marked as a maneuver disturbance window and its contribution to the overall risk value is reduced.

7. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 6, characterized in that, The modal confidence distributions for healthy and faulty states are constructed based on various characteristics. The confidence levels of these distributions are then corrected for modal integrity, noise level, and cross-modal consistency, and finally fused to obtain a comprehensive risk value. This includes: Modal confidence distributions for healthy state, mechanical imbalance fault, bearing deterioration fault, thermal anomaly fault, and electronic speed governor anomaly fault are generated for electrical characteristics, thermal characteristics, vibration characteristics, and control coupling characteristics, respectively. Based on the modal validity label, signal quality evaluation results, and consistency evaluation results among modal features, the credibility coefficient of each modality is obtained. After correcting the confidence distributions of each modality using the aforementioned confidence coefficients, the distributions are fused to obtain a comprehensive confidence distribution. The comprehensive risk value is determined based on the confidence level corresponding to the unhealthy state in the comprehensive confidence distribution, and the comprehensive risk value at multiple consecutive target times is stabilized.

8. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 7, characterized in that, The output of early warning, alarm, or protection control results based on the comprehensive risk value and the continuous exceeding of the threshold condition includes: A three-tiered risk threshold, comprising a first threshold, a second threshold, and a third threshold, is set within the range of 0 to 1; wherein the first threshold is lower than the second threshold, and the second threshold is lower than the third threshold. When the comprehensive risk value exceeds the first threshold for a continuous first period of time, an early warning result is output; When the comprehensive risk value exceeds the second threshold for a second consecutive period, an alarm result is output and a suggestion to reduce load or return to base is output. When the comprehensive risk value exceeds the third threshold for the third consecutive period, the protection control result is output. The protection control result includes at least one of the following: limiting speed, limiting maneuver range, performing redundancy switching, or outputting a forced landing command; Wherein, the second duration period is not less than the first duration period, and the third duration period is not greater than the second duration period.

9. The method for online monitoring and intelligent early warning of multiple parameters of a motor according to claim 8, characterized in that, Also includes: When the overall risk value is lower than the first threshold, the current aligned data frame and the current feature are used as healthy samples, and the feature mean and feature fluctuation range in the healthy baseline model are updated using the forgetting factor. When the comprehensive risk value reaches or exceeds the first threshold, the health baseline model update is frozen, and the aligned data frames, features, comprehensive risk value evolution sequence, control command evolution sequence, and warning level before and after the anomaly are encapsulated into an event data packet, which is then sent to a ground station, portable terminal, or cloud server for review and maintenance decision-making.

10. A multi-parameter online monitoring and intelligent early warning system for motors, characterized in that, include: The data access module is used to access the motor control command data, electronic speed governor telemetry data, vibration data, and temperature data of the flight controller; The time unification module is used to map motor control command data, electronic speed controller telemetry data, vibration data and temperature data with different sampling rates into aligned data frames and generate modal validity markers; The feature construction module is used to extract electrical features, rotational speed features, thermal features, vibration features, and control coupling features; The risk fusion module is used to construct modal confidence distributions, perform confidence corrections, and output a comprehensive risk value. The early warning decision module is used to output early warning, alarm or protection control results based on the comprehensive risk value and the continuous over-threshold condition; The baseline update and archiving module is used to update the health baseline model and generate event data packets in abnormal states.

Images