Dual-redundancy steering engine controller power supply monitoring method and system

Abstract

The present application relates to the technical field of aircraft rudder power supply health management, and specifically proposes a dual-redundancy rudder controller power supply monitoring method and system. The core of the method is to reversely use the power supply ripple "impact-misadjustment-spectrum" coupling effect: first, complete the hard fault rapid screening with the instantaneous impact deviation coefficient; then, extract the ripple peak valley misadjustment weight and wavelet packet singular spectrum entropy vector under the load matching condition, construct the "amplitude-spectrum" double feature field, and calculate the power supply degradation deviation with the help of the pre-established health sample library. The system divides the power supply state into four categories: healthy, mild degradation, moderate degradation and severe degradation, and outputs graded warning instructions through double threshold and PHM model criteria. The present application takes the ripple distortion evolution law as the basis for judgment, eliminates the influence of load, temperature drift and component tolerance, guarantees the sensitivity and robustness under complex avionics conditions, and greatly prolongs the service life of the rudder system.

Description

Technical Field

[0001] This invention relates to the field of aviation servo power health management technology, specifically to a method and system for monitoring the power supply of a dual-redundant servo controller. Background Technology

[0002] As a critical actuator in aircraft, the reliability of servo motor control directly impacts flight safety. Traditional dual-redundant servo controllers employ relay switching to achieve redundant power supply from two independent power sources. This presents a risk of power outages due to mechanical delays, and during switching, the drive circuit may fail to respond to servo commands due to power loss, potentially leading to control surface loss of control. In recent years, "OR logic" redundant power supply architectures have eliminated switching delays by using parallel power diodes, significantly improving power supply continuity and mission reliability. However, this architecture lacks sufficient detection capability for progressive and subtle soft faults, becoming a key bottleneck for the implementation of PHM (Prognostics and Detection Management) technology.

[0003] Aviation power supply failures are classified into two categories: hard failures and soft failures. Hard failures manifest as structural damage to components, exhibiting abrupt changes in characteristics and being easily identifiable. Soft failures, on the other hand, originate from non-destructive anomalies caused by component parameter drift, performance degradation, or the accumulation of thermal and electrical stress. They are characterized by gradual accumulation, weak initial symptoms, and multidimensional coupling, making them a key challenge for PHM (Prognostics and Maintenance Management). Under an "OR logic" architecture, when two power supplies experience soft failures simultaneously, the system can still maintain output without triggering a switch, but power quality and efficiency have substantially decreased. Without state awareness and remaining life prediction capabilities, this can easily lead to sluggish servo response and deteriorated control accuracy, potentially inducing flight quality degradation or even loss of control under extreme flight conditions. Existing monitoring technologies rely on static voltage threshold criteria and lack sensitive characterization and trend tracking methods for early degradation features such as ripple anomalies and dynamic consistency misalignment. This makes it difficult to achieve early diagnosis of soft failures, quantitative assessment of health status, and predictive maintenance decisions, thus hindering the comprehensive improvement of PHM capabilities in aviation electromechanical systems. Summary of the Invention

[0004] To address the technical challenges of early diagnosis of soft faults, quantitative assessment of health status, and predictive maintenance decisions in aircraft controller power supplies, this invention aims to provide a dual-redundant servo controller power supply monitoring method and system. The specific technical solution adopted is as follows: This invention proposes a power supply monitoring method for a dual-redundant servo controller, the method comprising: In controller power monitoring, the voltage and load power at different sampling times during the controller power monitoring cycle are acquired; based on the numerical characteristics of the voltage, it is determined whether a hard fault has occurred in the controller power supply. Based on the numerical distribution of load power during the monitoring period, a reference sample is obtained; the degree of deviation between the reference sample and the high and low voltage values ​​of the monitoring period is calculated to obtain the impact deviation coefficient of the monitoring period. Based on the voltage in the time sequence of the monitoring period, a voltage sampling sequence is obtained; the offset weight of voltage peaks and valleys is obtained by the degree of variation between the extreme value distributions in the voltage sampling sequence. The voltage sampling sequence is decomposed and singular spectral entropy analysis is performed to obtain the singular spectral entropy vector; based on the degree of dissimilarity between the reference sample and the singular spectral entropy vector of the monitoring period, the imbalance weight difference is weighted to obtain the waveform deviation of the monitoring period. The impact deviation coefficient, offset weight, and waveform deviation are sorted according to a fixed dimension to obtain a degradation feature sequence; the power supply status of the controller is determined based on the degradation feature sequence.

[0005] Furthermore, the method for determining hard faults includes: In the controller power supply monitoring, a preset hard fault double threshold is obtained; if the voltage sample value at N consecutive moments is greater than or equal to the upper threshold of the preset hard fault double threshold, or the voltage sample value at N consecutive moments is less than or equal to the lower threshold of the preset hard fault double threshold, it indicates that a hard fault has occurred in the controller power supply, where N represents the number of preset moments.

[0006] Furthermore, the method for obtaining the benchmark sample includes: By monitoring the power supply of unused controllers and collecting data on the monitoring cycle, a health cycle is obtained; and a health sample library is built by extracting the feature data of the health cycle. Calculate the absolute value of the difference between the load power of each health cycle and the monitoring cycle in the health sample library, select the electrical parameters of the top M health cycles with the smallest absolute difference, and obtain the benchmark sample, where M represents the preset number of health cycles.

[0007] Furthermore, the method for establishing the preset health sample library includes: Calculate the average of all voltages greater than the preset high voltage threshold within the healthy period to obtain the high voltage value; calculate the average of all voltages less than the preset high voltage threshold within the healthy period to obtain the low voltage value; construct a healthy sample library based on the load power, high voltage value, and low voltage value of the healthy period.

[0008] Furthermore, the method for obtaining the impact deviation coefficient includes: Based on the difference between the mean high voltage value of all benchmark samples and the high voltage value of the monitoring period, a deviation analysis is then performed to obtain the high voltage deviation. Based on the difference between the mean low voltage value of all benchmark samples and the low voltage value of the monitoring period, a deviation analysis is performed to obtain the low voltage deviation. The impact deviation coefficient of the monitoring period is obtained by fusing the characteristics of the high-level voltage deviation and the low-level voltage deviation.

[0009] Furthermore, the method for obtaining the imbalance weights includes: Based on the voltage in the time sequence of the monitoring period, obtain the voltage sampling sequence; based on the time distribution of extreme values ​​in the voltage sampling sequence, obtain the maximum value sequence and the minimum value sequence; calculate the feature fusion of the coefficient of variation of the maximum value sequence and the minimum value sequence to obtain the imbalance weight of the monitoring period.

[0010] Furthermore, the method for obtaining the singular spectral entropy vector includes: The voltage sampling sequence is decomposed into wavelet layers according to a preset number of decompositions to obtain the preset number of decomposition layers; the singular entropy spectrum of each decomposition layer is calculated and sorted according to the layer to obtain the singular spectral entropy vector of the voltage sampling sequence.

[0011] Furthermore, the method for obtaining the waveform deviation includes: The deviation of the imbalance weights in the monitoring period is obtained by comparing the imbalance weights in the monitoring period with the mean imbalance weights of the benchmark sample. The similarity between the singular spectral entropy vector of the monitoring period and the singular spectral entropy vector of all benchmark samples is obtained by inverting the similarity between the monitoring period and the benchmark samples. The waveform deviation during the monitoring period is obtained by weighted fusion between the imbalance weight deviation and the dissimilarity.

[0012] Furthermore, the method for determining the power status of the controller includes: The degradation feature sequence of the monitoring cycle is obtained by sorting and combining the impact deviation coefficient, offset weight and waveform deviation in a fixed order. Degradation feature sequences with different labels of healthy, mild degradation, moderate degradation and severe degradation are collected. The degradation feature sequences with different labels are trained by the feedforward neural network DFF to obtain the power degradation diagnosis model. The degradation feature sequence of the real-time monitoring cycle is input into the power degradation diagnosis model and the output is the controller power status label of the real-time monitoring cycle.

[0013] This invention offers the following advantages: It achieves seamless power supply for dual-redundant servo controllers by employing an OR logic redundant power supply architecture, eliminating the risk of power outage gaps associated with traditional relay switching and ensuring servo control continuity. Furthermore, it introduces a multi-dimensional feature analysis method based on ripple voltage, including impact deviation coefficient, offset weight, and waveform deviation, combined with a healthy sample library for load condition matching and comparison, significantly improving the early detection sensitivity for progressive soft faults such as component parameter drift and performance degradation. Finally, it constructs a "threshold filtering-model diagnosis" hierarchical early warning architecture, balancing computational resources and monitoring accuracy, providing low-latency, high-priority alarm responses to severely degraded states, achieving predictive health management, comprehensively improving the safety margin and maintainability of the power supply system, and effectively reducing the flight risk of servo control performance degradation due to power supply soft faults. Attached Figure Description

[0014] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart illustrating a power monitoring method for a dual-redundant servo controller according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a power supply circuit structure for a dual-redundant servo controller according to an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a dual-redundant servo controller power monitoring method and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] The specific scheme of the power monitoring method for a dual-redundant servo controller provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Please see Figure 1The diagram illustrates a flowchart of a power monitoring method for a dual-redundant servo controller according to an embodiment of the present invention.

[0020] Step S101: In the controller power monitoring, acquire electrical parameters at different sampling times during the controller power monitoring cycle, wherein the electrical parameters include voltage and load power; based on the numerical characteristics of the voltage, determine whether a hard fault has occurred in the controller power supply.

[0021] Dual-redundancy designs are generally divided into dual-redundancy designs for actuators and dual-redundancy designs for servo controllers. This invention mainly improves the traditional dual-redundancy design of two independent control circuits for the servo controller power supply. The improvement method is as follows: This application's dual-redundant servo controller power supply adopts an OR logic redundant power supply architecture. It uses two independent power supplies of the same model connected in parallel, employing four power diodes (A / B / C / D) to receive the positive and negative terminals of the power supply to the dual-redundant servo controller. The positive terminals of the two power supplies are connected to the anodes of diodes A and B, respectively, and the cathodes of diodes A and B are shorted to serve as the positive input power supply to the servo controller. The shorting point of the cathode lines of diodes A and B serves as the positive shorting point. The negative terminals of the two power supplies are connected to the cathodes of diodes C and D, respectively, and the anodes of diodes C and D are shorted to serve as the negative input power supply to the controller. The shorting point of the anode lines of diodes C and D serves as the negative shorting point. Figure 2 As shown, the load is the servo controller, power supply 1+ represents the positive terminal of the first power supply, power supply 2+ represents the positive terminal of the second power supply, power supply 1- represents the negative terminal of the first power supply, power supply 2- represents the negative terminal of the second power supply, and + represents the positive terminal and - represents the negative terminal.

[0022] In this application, for the power supply section of the dual-redundant servo controller, power diodes are connected in parallel to ensure that the servo controller can be powered when either power supply is normal, and the system will only have no output when both power supplies fail simultaneously; when one power supply fails, the other power supply can take over instantly and automatically, eliminating the power outage gap caused by the mechanical switching delay of traditional relays, and fundamentally ensuring the continuity of servo control and flight safety.

[0023] This application uses a high-speed ADC (Analog-to-Digital Converter) chip to acquire output data at the acquisition points between the positive shorting point and the positive terminal of the load, and transmits the data to the power monitoring module. The ADC uses dual-channel acquisition: channel 1 for voltage sampling and channel 2 for current sampling. The product of the voltage and current sampling values ​​at the same time is used as the load power at that moment.

[0024] A hard fault refers to structural damage or failure of certain critical components in a power supply system. It is usually sudden, and the electrical parameter symptoms before the fault are not obvious, but the electrical parameter characteristics at the time of the fault are clear. For overvoltage and undervoltage issues in the power supply of a servo controller experiencing a hard fault, the system checks whether there are N consecutive voltage sampling values ​​outside the preset dual thresholds for hard faults. If they are outside the range, it indicates a hard fault in the controller power supply.

[0025] It should be noted that N can be 10 in this embodiment.

[0026] Step S102: Based on the numerical distribution of load power during the monitoring period, obtain a reference sample; calculate the degree of deviation between the reference sample and the high and low voltage values ​​of the monitoring period to obtain the impact deviation coefficient of the monitoring period.

[0027] Soft faults in the servo controller power supply mainly originate from non-destructive anomalies caused by component parameter drift, performance degradation, or environmental factors. These faults reduce operating efficiency and output power quality, and are typically gradual, with initial fault characteristics not obvious, usually manifesting towards the end of the servo controller power supply's effective lifespan. Therefore, in the absence of hard faults, voltage sampling data is periodically sampled, with a period duration of 1 second and a sampling frequency of S×f, where f is the switching frequency of the power supply and S is the sampling rate, based on Shannon's sampling theorem. In this embodiment, S can be set to 50.

[0028] As a switching power supply, the DC output of the servo controller inevitably contains an AC component. This AC component, superimposed on the DC stability, is collectively referred to as ripple, and ripple voltage has a high fault identification rate. During the monitoring period of the servo controller power supply, the voltage samples are arranged in ascending order. The average of the first 25% of the sampled values ​​is recorded as the low-order voltage value, and the average of the last 25% is recorded as the high-order voltage value.

[0029] The voltage sampling value of the servo controller's power supply is related to the load power, the degree of power supply degradation, and the usage time. Therefore, by periodically sampling the power supply of the new servo controller in its healthy state, the load power of the cycle is obtained. Based on the difference in the distribution of the load power values ​​between the periodically sampled load power in the healthy state and the load power values ​​in the current monitoring cycle, a benchmark sample is obtained. Since the performance degradation of the switching power supply is often accompanied by an increase in output ripple, the degree of voltage distortion can be analyzed through the voltage operating status to identify the degradation state of the controller power supply. Therefore, by weighted fusion of the deviations of the high and low voltage values ​​in the current monitoring cycle and the benchmark sample, the impact deviation coefficient of the monitoring cycle is obtained. The larger the current impact deviation coefficient, the more severe the voltage distortion and the more significant the power quality degradation.

[0030] Step S103: Obtain the voltage sampling sequence based on the voltage in the monitoring cycle time sequence; obtain the voltage peak and valley offset weights by the degree of variation between the extreme value distributions in the voltage sampling sequence.

[0031] The voltage samples within the monitoring period are sorted chronologically to obtain a voltage sampling sequence, which characterizes the operating status of the controller power supply voltage within the monitoring period. When the power supply to the servo controller experiences performance degradation (soft fault), the parameters of internal key components deviate from their optimal design values, leading to a decrease in the dynamic balance capability, stability, and consistency of the power supply system. This state causes irregular fluctuations in the ripple amplitude superimposed on the output DC voltage, and may even trigger occasional, random voltage spikes or ringing, significantly increasing peak values ​​and decreasing trough values ​​in the voltage sampling sequence. Therefore, the degree of variation between extreme value distributions is extracted to obtain the offset weight of voltage peaks and troughs. A larger offset weight indicates poorer ripple periodicity consistency, more frequent voltage spikes or ringing, reflecting a degradation in the power supply's dynamic balance capability, and demonstrating high sensitivity to minor soft faults.

[0032] Step S104: Decompose the voltage sampling sequence and perform singular spectral entropy analysis to obtain the singular spectral entropy vector; based on the degree of dissimilarity between the reference sample and the singular spectral entropy vector of the monitoring period, perform weighted processing of the misalignment weight difference to obtain the waveform deviation of the monitoring period.

[0033] When the power quality of the controller power supply deteriorates, the power fluctuations of distributed energy sources and the frequent charging and discharging of the energy storage system lead to a more complex frequency domain energy distribution structure of the voltage waveform, irregular superposition of harmonic components, and may even trigger occasional, random voltage flicker or distortion, significantly altering the energy distribution characteristics of the voltage sampling sequence across the entire frequency band. Therefore, a decomposition algorithm can be used to decompose the voltage sampling sequence into n sub-bands, calculate the singular entropy spectrum corresponding to the decomposition coefficients of each sub-band, and construct a singular spectral entropy vector to characterize the complexity or irregularity of the voltage waveform in the frequency domain. Furthermore, the dissimilarity between the singular spectral entropy vector of the monitoring period and the benchmark sample is calculated, and weighted fusion is performed using the relative deviation of the offset weights to obtain the waveform deviation of the monitoring period. A larger waveform deviation indicates a more significant difference between the voltage waveform energy distribution structure and the healthy benchmark, higher frequency domain complexity, and a more severe hidden degradation of power quality, demonstrating high identification accuracy for highly hidden, mild degradation features. It should be noted that n represents the preset decomposition number, which can be 8 in this embodiment.

[0034] Step S105: Sort the impact deviation coefficient, offset weight and waveform deviation according to a fixed dimension to obtain a degradation feature sequence; determine the controller power supply status based on the degradation feature sequence.

[0035] Degradation feature sequences are obtained by arranging them in the order of impact deviation coefficient, offset weight, and waveform deviation, and are used to characterize the operating status of the controller power supply during the monitoring period. Degradation feature sequences of healthy, mild, moderate, and severe degradation states are collected under laboratory or actual operating conditions. These degradation feature sequences from different operating conditions are used as inputs to a feedforward neural network (DFF) to train the power supply degradation diagnostic model. Furthermore, a dual degradation threshold and the power supply degradation diagnostic model are used to determine the state of the controller power supply.

[0036] This invention achieves seamless power supply for dual-redundant servo controllers by employing an OR logic redundant power supply architecture, eliminating the risk of power outage gaps associated with traditional relay switching and ensuring continuous servo control. It introduces a multi-dimensional feature analysis method based on ripple voltage, including impact deviation coefficient, offset weight, and waveform deviation, combined with a healthy sample library for load condition matching and comparison, significantly improving the early detection sensitivity for progressive soft faults such as component parameter drift and performance degradation. Furthermore, it constructs a "threshold filtering-model diagnosis" hierarchical early warning architecture, balancing computational resources and monitoring accuracy, providing low-latency, high-priority alarm responses to severely degraded states, achieving predictive health management, comprehensively improving the safety margin and maintainability of the power supply system, and effectively reducing the flight risk of servo control performance degradation due to power supply soft faults.

[0037] Preferably, in some possible implementations of the embodiments of the present invention, the method for determining hard faults includes: In controller power supply monitoring, preset hard fault dual thresholds are obtained. If the voltage sample value at N consecutive time points is greater than or equal to the upper threshold of the preset hard fault dual thresholds, or if the voltage sample value at N consecutive time points is less than or equal to the lower threshold of the preset hard fault dual thresholds, it indicates that a hard fault has occurred in the controller power supply. By determining the thresholds at consecutive time points, transient interference glitches can be eliminated, effectively distinguishing between real hard faults and occasional noise, avoiding false alarms or missed detections caused by single-point misjudgments, significantly improving the reliability and real-time performance of hard fault identification, and providing the first line of defense for the power supply safety of the servo controller.

[0038] It should be noted that: in this embodiment, the upper threshold of the preset hard fault dual threshold is set to 130% of the rated output voltage; the lower threshold is set to 70% of the rated output voltage; N represents the preset number of time moments, which is 10 in this embodiment.

[0039] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the benchmark sample includes: A health cycle is obtained by collecting electrical parameters from unused controller power supplies during monitoring cycles. Feature data of the electrical parameters in the health cycles, including high and low voltage values ​​and load power, is extracted. Feature data from all health cycles are collected to establish a health sample library. The absolute value of the difference between the load power of each health cycle and the monitoring cycle in the health sample library is calculated. The feature data of the M health cycles with the smallest absolute difference are selected to obtain M benchmark samples. These samples characterize the state of the power supply during healthy operation under load conditions most similar to the current control, and are used to eliminate the interference of load fluctuations on degradation assessment. In this embodiment, the number of health cycles can be 3600, and M can be 10.

[0040] Preferably, in some possible implementations of the embodiments of the present invention, the method for establishing a health sample bank includes: The average value of all voltages greater than the preset high voltage threshold is calculated in each health cycle to obtain the high voltage value; the average value of all voltages less than the preset high voltage threshold is calculated in each health cycle to obtain the low voltage value; a health sample library is constructed based on the load power, high voltage value and low voltage value of the health cycle. The health sample library represents the baseline operating characteristics of the new servo controller power supply in a healthy state, eliminating the interference of load fluctuations on degradation assessment.

[0041] It should be noted that in this embodiment, the preset low voltage threshold and the preset high voltage threshold are the upper and lower quartiles of the healthy cycle, respectively. The upper and lower quartiles are obtained by arranging the voltage sampling values ​​of the healthy cycle from smallest to largest, with the 25th percentile voltage sampling value being the lower quartile and the 75th percentile voltage sampling value being the upper quartile.

[0042] Preferably, in some possible implementations of the embodiments of the present invention, the impact deviation coefficient is calculated as follows: In an OR logic redundant power supply architecture, ripple voltage surges are susceptible to load fluctuations and component tolerance interference. By employing load matching conditions and using a benchmark sample for normalization analysis, operating condition differences can be eliminated, degradation characteristics highlighted, and the degree of deviation of the ripple voltage amplitude from the healthy benchmark can be represented, providing a dimensionless indicator for subsequent soft fault diagnosis and graded early warning. Therefore, by weighted fusion of the relative deviations of high-level and low-level voltage deviations, the surge deviation coefficient for that cycle is obtained. The larger the surge deviation coefficient, the more significant the deviation of the ripple voltage surge and undersurge amplitude from the healthy benchmark, reflecting the degree of power quality degradation. In a specific implementation of this invention, the surge deviation coefficient is expressed by the formula: In the formula, This represents the impact deviation coefficient for the current monitoring period. , These represent the high-order and low-order voltage values ​​for the current monitoring period, respectively. , This represents the mean of the higher-order voltage values ​​and the mean of the lower-order voltage values ​​of all reference samples. Indicates the absolute value sign. , These represent the high-level voltage deviation and the low-level voltage deviation, respectively.

[0043] It should be noted that when the mean of the high-level voltage value or the mean of the low-level voltage value is 0, the corresponding deviation of the high-level voltage and the deviation of the low-level voltage are both 0.

[0044] In another embodiment, the impact deviation coefficient is obtained by multiplying the high-level voltage deviation by the low-level voltage deviation.

[0045] Preferably, in some possible implementations of the embodiments of the present invention, the imbalance weight is calculated as follows: In the early stages of power supply performance degradation, the dispersion of ripple extreme values ​​is easily affected by random noise and transient disturbances. The variation state of the extreme value sequence can represent the degree of degradation of the peak-valley period stability of the ripple voltage. Therefore, the voltage sampling sequence is used as the input of the Automatic multiscale-based peak detection (AMPD) algorithm to obtain the maximum and minimum value sequences in the voltage sampling sequence. By fusing the characteristic of the coefficient of variation of the maximum and minimum value sequences, the offset weight of the monitoring period is obtained. The larger the offset weight, the more severe the dispersion of the ripple peaks and troughs and the worse the period consistency, which can reflect the offset characteristics of the power supply's dynamic balance capability degradation. The calculation of the multiscale peak and coefficient of variation is common knowledge, and the specific calculation steps will not be elaborated here. In a specific implementation of this invention, the offset weight is expressed by the formula: In the formula, This indicates the imbalance weight for the current monitoring period. , These represent the coefficients of variation for the maximum and minimum value sequences in the voltage sampling sequence, respectively.

[0046] In another embodiment, the imbalance weights are obtained by multiplying the coefficients of variation of the maximum and minimum sequences.

[0047] Preferably, in some possible implementations of the embodiments of the present invention, the singular spectral entropy vector is calculated as follows: When power quality deteriorates covertly, the disordered frequency domain energy structure is easily masked by the superposition of fundamental frequency components and harmonics. Wavelet packet decomposition, using multi-scale sub-bands as units, is employed to calculate singular spectral entropy. This method can remove time-domain aliasing, highlight frequency domain complexity, and represent the irregularity of the energy distribution across the entire frequency band of the voltage waveform. Therefore, by sorting and combining the singular entropy spectra of each decomposition layer according to frequency hierarchy, the singular spectral entropy vector for that period is obtained. Higher singular spectral entropy values ​​in each dimension of the vector indicate a more complex frequency domain energy structure of the voltage waveform in that dimension, a more significant difference from the healthy baseline, and characterize the spectral features of the covert degradation of power quality. The calculation of singular spectral entropy is common knowledge, and the specific calculation steps will not be elaborated here.

[0048] Preferably, in some possible implementations of the embodiments of the present invention, the waveform deviation is calculated as follows: During the mild degradation stage of the power supply, single features are easily affected by load matching deviations and temperature drift. A weighted fusion of dual features, coupled with analysis based on offset weight deviation and singular spectrum dissimilarity, can compensate for local sensitivity deficiencies and highlight global waveform anomalies. This represents the health benchmark degree of the comprehensive deviation of ripple voltage in the time and frequency domains, providing a multi-dimensional quantitative indicator for subsequent neural network state classification and predictive maintenance decisions. Therefore, by weighting and fusing the offset weight deviation and singular spectrum entropy vector dissimilarity between the monitoring period and the benchmark sample, the waveform deviation for that period is obtained. The larger the waveform deviation, the more significant the superposition effect of ripple peak-valley consistency misalignment and frequency domain energy distribution anomalies. In a specific implementation of this invention, the waveform deviation is expressed by the formula: In the formula, This indicates the waveform deviation during the current monitoring period. This indicates the imbalance weight for the current monitoring period. This represents the mean of the imbalance weights for all baseline samples. This represents the singular spectral entropy vector for the current monitoring period. This represents the singular spectral entropy vector of the m-th reference sample. This represents the normalization function; in this embodiment, the sigmoid function is used. Represents the cosine similarity function. This indicates the dissimilarity between the monitoring period and the baseline sample. This indicates the degree of imbalance weight deviation during the monitoring period. This represents a constant parameter to prevent the denominator from being 0; in this embodiment, it is set to 0.001.

[0049] In another embodiment, the imbalance weight deviation of the monitoring period is calculated as follows: The dissimilarity between the monitoring period and the benchmark sample is calculated by using the DTW distance (Dynamic Time Warping) between the singular spectral entropy vectors of the monitoring period and the benchmark sample.

[0050] Preferably, in some possible implementations of the embodiments of the present invention, the method for determining the power state of the controller is as follows: Under laboratory conditions or actual operating conditions, voltage and current data of the servo controller power supply throughout its entire lifecycle, from healthy to failure, are acquired. Degradation feature vectors are obtained through the steps described above and manually labeled to construct a training set. The label categories are: healthy, mild degradation, moderate degradation, and severe degradation. A feedforward neural network (DFF) is selected to establish a power supply degradation diagnostic model. The model is trained using the training dataset. The DFF structure consists of an input layer, two hidden layers, and an output layer. All layers are fully connected. The Adam gradient descent algorithm is used for learning and optimization. The initial learning rate is set to 0.002, the model size to 100, the number of training iterations to 500, and the activation function to ReLU.

[0051] The degradation feature sequence is obtained by sorting the parameters in the order of impact deviation coefficient, offset weight, and waveform deviation, which can characterize the power supply status of the servo controller during operation. The impact deviation coefficient in the degradation feature sequence has high detection accuracy for impact amplitude deviation characteristics caused by severe power supply degradation. Furthermore, severe power supply degradation is the soft fault that has the greatest impact on the normal operation of the servo controller. To ensure the safety of aircraft and servo operation, it is necessary to promptly alert management personnel. This application sets dual thresholds in the following manner. , This enables refined monitoring and tiered early warning of soft faults in the power supply of the servo controller. (1) Obtain the impact deviation coefficient of the servo controller power supply during the monitoring period. If the impact deviation coefficient exceeds the degradation threshold... The output result is severe degradation, and the power monitoring module issues a level two alarm to remind the management personnel to check the power supply in time.

[0052] (2) If the impact deviation coefficient is higher than the degradation threshold However, it did not exceed the degradation threshold. Based on the above steps, the degradation feature vector of the monitoring period is obtained and used as the input of the power degradation diagnosis model. The output result is obtained. If the output result is severe degradation, the power monitoring module issues a level 2 alarm to remind the management personnel to check the power supply in time. If the output result is moderate degradation, the power monitoring module issues a level 3 alarm to provide a forward warning for the power supply of the servo controller. If the output result is normal (i.e., mild degradation or healthy), the power monitoring module does not issue an alarm.

[0053] (3) If the impact deviation coefficient does not exceed The output shows a normal state, and the power monitoring module does not issue an alarm.

[0054] It should be noted that: degradation threshold The calculation method is as follows: obtain the training dataset of the power degradation diagnostic model, calculate the mean of the shock deviation coefficient of all training samples labeled as severely degraded; degradation threshold. The calculation method is to calculate the mean of the shock deviation coefficient of all training samples labeled as mild or moderate degradation.

[0055] Thus, a power monitoring method for a dual-redundant servo controller is realized.

[0056] Based on the same inventive concept as the above methods, this application also provides a dual-redundant servo controller power monitoring system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the above methods.

[0057] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0058] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A power supply monitoring method for a dual-redundant servo controller, characterized in that, The method includes: In controller power monitoring, the voltage and load power at different sampling times during the controller power monitoring cycle are acquired; based on the numerical characteristics of the voltage, it is determined whether a hard fault has occurred in the controller power supply. Based on the numerical distribution of load power during the monitoring period, a reference sample is obtained; the degree of deviation between the reference sample and the high and low voltage values ​​of the monitoring period is calculated to obtain the impact deviation coefficient of the monitoring period. Based on the voltage in the time sequence of the monitoring period, a voltage sampling sequence is obtained; the offset weight of voltage peaks and valleys is obtained by the degree of variation between the extreme value distributions in the voltage sampling sequence. The voltage sampling sequence is decomposed and singular spectral entropy analysis is performed to obtain the singular spectral entropy vector; based on the degree of dissimilarity between the reference sample and the singular spectral entropy vector of the monitoring period, the imbalance weight difference is weighted to obtain the waveform deviation of the monitoring period. The impact deviation coefficient, offset weight, and waveform deviation are sorted according to a fixed dimension to obtain a degradation feature sequence; the controller power supply status is determined based on the degradation feature sequence. The method for obtaining the benchmark sample includes: By monitoring the power supply of unused controllers and collecting data on the monitoring cycle, a health cycle is obtained; and a health sample library is built by extracting the feature data of the health cycle. Calculate the absolute value of the difference between the load power of each health cycle and the monitoring cycle in the health sample library, select the electrical parameters of the top M health cycles with the smallest absolute difference, and obtain the benchmark sample, where M represents the preset number of health cycles; The method for establishing the preset health sample library includes: Calculate the average of all voltages greater than the preset high voltage threshold within the healthy period to obtain the high voltage value; calculate the average of all voltages less than the preset high voltage threshold within the healthy period to obtain the low voltage value; construct a healthy sample library based on the load power, high voltage value, and low voltage value of the healthy period.

2. The power supply monitoring method for a dual-redundant servo controller according to claim 1, characterized in that, The methods for determining hard faults include: In the controller power supply monitoring, a preset hard fault double threshold is obtained; if the voltage sample value at N consecutive moments is greater than or equal to the upper threshold of the preset hard fault double threshold, or the voltage sample value at N consecutive moments is less than or equal to the lower threshold of the preset hard fault double threshold, it indicates that a hard fault has occurred in the controller power supply, where N represents the number of preset moments.

3. The power supply monitoring method for a dual-redundant servo controller according to claim 1, characterized in that, The method for obtaining the impact deviation coefficient includes: Based on the difference between the mean high voltage value of all benchmark samples and the high voltage value of the monitoring period, a deviation analysis is then performed to obtain the high voltage deviation. Based on the difference between the mean low voltage value of all benchmark samples and the low voltage value of the monitoring period, a deviation analysis is performed to obtain the low voltage deviation. The impact deviation coefficient of the monitoring period is obtained by fusing the characteristics of the high-level voltage deviation and the low-level voltage deviation.

4. The power supply monitoring method for a dual-redundant servo controller according to claim 1, characterized in that, The method for obtaining the imbalance weights includes: Based on the voltage in the time sequence of the monitoring period, obtain the voltage sampling sequence; based on the time distribution of extreme values ​​in the voltage sampling sequence, obtain the maximum value sequence and the minimum value sequence; calculate the feature fusion of the coefficient of variation of the maximum value sequence and the minimum value sequence to obtain the imbalance weight of the monitoring period.

5. The power supply monitoring method for a dual-redundant servo controller according to claim 1, characterized in that, The method for obtaining the singular spectral entropy vector includes: The voltage sampling sequence is decomposed into wavelet layers according to a preset number of decompositions to obtain the preset number of decomposition layers; the singular entropy spectrum of each decomposition layer is calculated and sorted according to the layer to obtain the singular spectral entropy vector of the voltage sampling sequence.

6. The power supply monitoring method for a dual-redundant servo controller according to claim 1, characterized in that, The method for obtaining the waveform deviation includes: The deviation of the imbalance weights in the monitoring period is obtained by comparing the difference between the imbalance weights in the monitoring period and the mean imbalance weights of the benchmark sample. The similarity between the singular spectral entropy vector of the monitoring period and the singular spectral entropy vector of all benchmark samples is obtained by inverting the similarity between the monitoring period and the benchmark samples. The waveform deviation during the monitoring period is obtained by weighted fusion between the imbalance weight deviation and the dissimilarity.

7. The power supply monitoring method for a dual-redundant servo controller according to claim 1, characterized in that, The method for determining the power status of the controller includes: The degradation feature sequence of the monitoring cycle is obtained by sorting and combining the impact deviation coefficient, offset weight and waveform deviation in a fixed order. Degradation feature sequences with different labels of healthy, mild degradation, moderate degradation and severe degradation are collected. The degradation feature sequences with different labels are trained by the feedforward neural network DFF to obtain the power degradation diagnosis model. The degradation feature sequence of the real-time monitoring cycle is input into the power degradation diagnosis model and the output is the controller power status label of the real-time monitoring cycle.

8. A power supply monitoring system for a dual-redundant servo controller, characterized in that, The system includes a processor and a memory; wherein the memory stores a computer program that can run on the processor; and the processor executes the program stored in the memory to implement the steps of the power monitoring method for a dual-redundant servo controller as described in any one of claims 1-7.

Images