A method for identifying the health status of joint motors with oil ingress faults by integrating current timing and low-frequency power ratio.

By integrating current timing and low-frequency power ratio, the problem of difficulty in distinguishing oil ingress faults in industrial robot joint motors in existing technologies has been solved, enabling accurate identification and stable monitoring of fault levels.

CN123069063APending Publication Date: 2026-11-10SOUTHEAST UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611231701.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing health monitoring methods are difficult to effectively distinguish different degrees of oil ingress faults in the joint motors of industrial robots. The original current waveform is easily affected by noise and load changes, and the low-frequency power ratio loses high-frequency local details when used alone.

Method used

The method of fusing current timing and low-frequency power ratio constructs a health characterization of the original current-low-frequency power ratio through Clarke transform, Park transform, narrowband filtering and Hilbert envelope extraction, preserving local waveform details and enhancing fault low-frequency modulation information.

Benefits of technology

It improves the ability to distinguish between different fault levels, enhances the accuracy and stability of health status identification, and is suitable for industrial robot condition monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069063A_ABST
    Figure CN123069063A_ABST
Patent Text Reader

Abstract

This invention discloses a method for identifying the health status of an articulated motor with oil inlet faults by fusing current time series data and low-frequency power ratio. The method takes current time series data as input, preprocesses it to obtain a raw current sample sequence, then processes and extracts raw current features from the raw current sample sequence. Bandpass filtering, Hilbert envelope extraction, trend-fluctuation decomposition, relative modulation sequence construction, downsampling, power spectrum estimation, and low-frequency power ratio calculation are then performed on the current time series data to obtain a low-frequency power ratio sequence. The raw current features and the low-frequency power ratio sequence are processed and fused to construct a raw current-low-frequency power ratio fused health characterization. The health status identification result corresponding to the severity of the oil inlet fault is output through this raw current-low-frequency power ratio fused health characterization. This method effectively preserves the linear processing flow from raw current to low-frequency power ratio features and then to convolutional neural network classification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of industrial robot condition monitoring and fault diagnosis technology, specifically involving a method for identifying the health status of joint motors with oil ingress faults by integrating current timing and low-frequency power ratio. Background Technology

[0002] When industrial robots perform repetitive tasks, their joint current signals exhibit a clear periodicity. However, the variable start-stop duration, disturbance noise, and load fluctuations can blur the periodic boundaries, making it difficult to directly use the raw signals for health monitoring modeling. Compared to simulation conditions, industrial field current signals are also affected by factors such as acquisition link errors, environmental noise, drive disturbances, load fluctuations, and limited sample size, resulting in a more complex distribution and weaker fault characteristics. For oil ingress faults in permanent magnet synchronous motors, the evolution in the current observation domain follows a pattern of "baseline rise – enhanced fluctuation." As the severity of the oil ingress fault increases, the relative low-frequency correlation energy of the current envelope in the modulation sequence gradually increases, and the low-frequency power ratio index value rises accordingly.

[0003] Existing health monitoring methods typically use statistical quantities such as raw current waveform, mean, variance, peak value, or frequency domain energy for state characterization. While raw current waveforms can preserve local waveform details and transient changes, they are easily affected by operating cycle time, load variations, periodic boundaries, and noise disturbances. Although single statistical indicators are easy to interpret, they often fail to simultaneously characterize the trend changes and low-frequency fluctuations caused by oil inlet faults. Low-frequency power ratios can highlight fault-related low-frequency modulation information, but if used as a sole input, they may lose high-frequency local details and boundary transition information in the raw current, affecting the differentiation between adjacent fault levels.

[0004] Therefore, it is necessary to construct a dual-source input health monitoring process based on the original current signal and the low-frequency power ratio index sequence, so that the local waveform details of the original current and the mechanism enhancement information of the low-frequency power ratio complement each other, thereby improving the ability to distinguish samples with different fault levels. Summary of the Invention

[0005] The purpose of this invention is to provide a method for identifying the health status of an articulated motor with oil ingress faults by integrating current timing and low-frequency power ratio. This method focuses on the baseline rise and enhanced fluctuation patterns exhibited by oil ingress faults in the current observation domain. While preserving the details of the original current local waveform, it introduces a low-frequency power ratio sequence as a mechanism enhancement branch to construct a fused health characterization of the original current and low-frequency power ratio, thereby improving the ability to distinguish health statuses at adjacent fault levels.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for identifying the health status of a joint motor with oil ingress faults by integrating current timing and low-frequency power ratio includes the following steps:

[0008] S1: Obtain the three-phase stator current timing data of the servo joint motor to be monitored, and perform period alignment, sample segmentation and length unification on the three-phase stator current timing data to obtain the original current sample sequence;

[0009] S2: Scale normalize the original current sample sequence and extract the original current features used to characterize local waveform morphology, transient changes and fine-grained temporal texture;

[0010] S3: Perform Clarke and Park transformations on the three-phase stator current timing data to obtain the orthogonal current components in the two-phase stationary coordinate system and the current components in the synchronous rotating coordinate system, and construct a complex vector of current space based on the orthogonal current components.

[0011] S4: Narrowband filtering is performed using the main conduction angular frequency of the current space complex vector as the center frequency to obtain the narrowband current signal in the fundamental frequency neighborhood;

[0012] S5: Perform Hilbert envelope extraction on the narrowband current signal to obtain the amplitude envelope sequence, and perform trend-fluctuation decomposition on the amplitude envelope sequence to extract the trend component and fluctuation component;

[0013] S6: Construct a relative modulation sequence based on the trend component and the fluctuation component, calculate the power spectrum of the relative modulation sequence, calculate the low-frequency power and high-frequency power based on the fault-related cutoff frequency, and calculate the low-frequency power ratio sequence.

[0014] S7: The original current features obtained in step S2 and the low-frequency power ratio sequence obtained in step S6 are aligned in time length, normalized in scale, and fused to construct the original current-low-frequency power ratio fused health characterization.

[0015] S8: Based on the original current-low frequency power ratio fusion health characterization, perform health status identification and output the health status identification result corresponding to the severity of the oil inlet fault.

[0016] As a preferred technical solution of the present invention: In step S1, the period alignment includes dividing the continuous three-phase stator current time sequence data into periodic samples of uniform length according to the working cycle, so that samples of different health states are comparable in terms of time scale and sample length.

[0017] As a preferred technical solution of the present invention: In step S3, the three-phase stator current timing data , and Perform Clarke transform to obtain Orthogonal current components of the axis and Orthogonal current components of the axis :

[0018] (1);

[0019] The complex vector of current space is calculated. :

[0020] (2);

[0021] in, It is the symbol for the imaginary unit in complex numbers.

[0022] As a preferred technical solution of the present invention: In step S4, the cutoff frequency of the bandpass filter of the narrowband filter is... and upper cutoff frequency Determine using the following method:

[0023] First, define the lower bound for removing DC and slow-varying trends. And the upper limit for retaining the main modulation component of the current. ;

[0024] Further order:

[0025] (3);

[0026] (4);

[0027] in To ensure the upper cutoff frequency Greater than the lower cutoff frequency Positive numbers, This is below the safety limit for Nyquist frequencies.

[0028] As a preferred technical solution of the present invention: in step S5, the amplitude envelope sequence is obtained from an analytical signal, the analytical signal... for:

[0029] (5);

[0030] in, It is a narrowband current signal. For Hilbert transform,

[0031] Amplitude envelope for:

[0032] (6).

[0033] As a preferred technical solution of the present invention: in step S5, the trend component Obtained from the moving average operator:

[0034] (7);

[0035] in, It is a discrete amplitude envelope sequence. This represents the value of the discrete amplitude envelope sequence at the (n - q)th sampling point. The total number of samples in the discrete amplitude envelope sequence;

[0036] The fluctuation component The calculation is as follows:

[0037] (8);

[0038] (9);

[0039] in, It is a discrete amplitude envelope sequence. , T represents the fluctuation component of the discrete amplitude envelope sequence at the nth sampling point. s The sampling interval is... The length of the sliding window. The cutoff frequency is The low-pass filter operator.

[0040] As a preferred embodiment of the present invention: in step S6, the relative modulation sequence is:

[0041] (10);

[0042] in, For fluctuation components, As a trend component, This is a regularization constant used to avoid numerical instability caused by excessively small local baselines.

[0043] As a preferred technical solution of the present invention: in step S6, the relative modulation sequence is downsampled to prevent aliasing before power spectrum estimation to obtain the downsampled relative modulation sequence. and in the The power spectral density is estimated using the Welch method within each analysis window. ,in, For discrete frequency sampling point index, The highest frequency sampling point number corresponding to the upper limit of the analysis frequency;

[0044] The fault-related low-frequency power is:

[0045] (11);

[0046] The fault-related high-frequency power is:

[0047] (12);

[0048] in, This refers to the frequency sampling point number corresponding to the low-frequency cutoff frequency. Frequency resolution;

[0049] The low-frequency power ratio sequence is then:

[0050] (13);

[0051] in, is the power regularization constant.

[0052] As a preferred technical solution of the present invention: In step S7, let the first... The original current characteristics of each period sample are as follows The low-frequency power ratio sequence is Scale normalization is performed separately:

[0053] (14);

[0054] (15);

[0055] in , , and The mean and standard deviation are obtained from the training samples. , It is a numerically stable term;

[0056] The original current-low-frequency power ratio combined health characterization is as follows:

[0057] (16);

[0058] in This is for sequential splicing.

[0059] As a preferred technical solution of the present invention: in step S8, the health status identification result includes at least one of normal status, minor fault, secondary serious fault, more serious fault or serious fault.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] (1) Starting from the three-phase stator current time sequence data, this invention maps the current modulation phenomenon caused by the oil inlet fault of the servo joint to the envelope domain through Clarke transform, Park transform, narrowband complex vector construction and Hilbert envelope extraction, thereby improving the observability of fault-related information.

[0062] (2) The present invention constructs a relative modulation sequence by trend-fluctuation decomposition, so that the intensity of dynamic fluctuation is expressed with reference to the local baseline, which can weaken the influence of amplitude scale differences in different operating stages on health characterization.

[0063] (3) The present invention integrates the low-frequency power ratio as mechanism enhancement information with the original current characterization, which avoids the problem of losing local timing details by using the low-frequency power ratio alone, and also enhances the ability of the original current alone to express the low-frequency modulation law of the fault.

[0064] (4) The original current-low frequency power ratio fusion health characterization constructed in this invention can be used as the input for subsequent health status identification, which can improve the model's ability to distinguish samples with different fault levels and has good engineering applicability. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of the overall process of the method of the present invention.

[0066] Figure 2 This is a schematic diagram of the low-frequency power ratio feature extraction process in this invention.

[0067] Figure 3 This is a schematic diagram showing the current waveform and low-frequency power ratio characteristics of normal and faulty samples in this invention. Detailed Implementation

[0068] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0069] like Figure 1 As shown, the method for identifying the health status of a joint motor with oil ingress fault, which integrates current timing and low-frequency power ratio, proposed in this invention, includes the following steps:

[0070] S1: Acquire the three-phase stator current timing data of the servo joint motor to be monitored, and perform period alignment, sample segmentation and length unification on the three-phase stator current timing data to obtain the original current sample sequence. The original current sample sequence is used to preserve the local waveform morphology, transient changes and fine-grained timing texture, as one source of information for fusing health characterization.

[0071] The period alignment includes dividing the continuous three-phase stator current time series data into period samples of uniform length according to the operating cycle, so that samples of different health states are comparable in terms of time scale and sample length.

[0072] S2: Scale normalize the original current sample sequence and extract the original current features used to characterize local waveform morphology, transient changes and fine-grained temporal texture;

[0073] S3: Perform Clarke and Park transformations on the three-phase stator current timing data to obtain the orthogonal current components in the two-phase stationary coordinate system and the current components in the synchronous rotating coordinate system, and construct a complex vector of current space based on the orthogonal current components.

[0074] Among them, the three-phase stator current timing data , and Perform Clarke transform to obtain Orthogonal current components of the axis and Orthogonal current components of the axis :

[0075] (1);

[0076] The complex vector of current space is calculated. :

[0077] (2);

[0078] in, It is the symbol for the imaginary unit in complex numbers.

[0079] Narrowband filtering is performed based on the principal conduction angular frequency of the complex vector in the current space to extract the narrowband current signal in the neighborhood of the fundamental frequency. The narrowband current signal can be regarded as a signal with slowly varying amplitude and superimposed low-frequency modulation within the local time window, and therefore can be used to extract the envelope features reflecting the fault evolution.

[0080] S4: Narrowband filtering is performed using the main conduction angular frequency of the current space complex vector as the center frequency to obtain the narrowband current signal in the fundamental frequency neighborhood;

[0081] Among them, the cutoff frequency of the bandpass filter in narrowband filtering and upper cutoff frequency Determine using the following method:

[0082] First, define the lower bound for removing DC and slow-varying trends. And the upper limit for retaining the main modulation component of the current. ;

[0083] Further order:

[0084] (3);

[0085] (4);

[0086] in To ensure the upper cutoff frequency Greater than the lower cutoff frequency Positive numbers, To ensure safety below the Nyquist frequency limit, in embodiments of the present invention, Hz, Hz.

[0087] S5: Perform Hilbert envelope extraction on the narrowband current signal to obtain the amplitude envelope sequence, and perform trend-fluctuation decomposition on the amplitude envelope sequence to extract the trend component and fluctuation component;

[0088] The amplitude envelope sequence is obtained from an analytic signal, and the analytic signal... for:

[0089] (5);

[0090] in, It is a narrowband current signal. For Hilbert transform,

[0091] Amplitude envelope for:

[0092] (6).

[0093] The trend component Obtained from the moving average operator:

[0094] (7);

[0095] in, It is a discrete amplitude envelope sequence. This represents the value of the discrete amplitude envelope sequence at the (n - q)th sampling point. The total number of samples in the discrete amplitude envelope sequence;

[0096] The fluctuation component The calculation is as follows:

[0097] (8);

[0098] (9);

[0099] in, It is a discrete amplitude envelope sequence. , T represents the fluctuation component of the discrete amplitude envelope sequence at the nth sampling point. s The sampling interval is... The length of the sliding window. The cutoff frequency is The low-pass filter operator.

[0100] The trend component reflects the local baseline level, while the fluctuation component reflects the dynamic fluctuations relative to the baseline.

[0101] To mitigate amplitude scale differences across different operating phases and local windows, a relative modulation sequence is further constructed.

[0102] S6: Construct a relative modulation sequence based on the trend component and the fluctuation component, calculate the power spectrum of the relative modulation sequence, calculate the low-frequency power and high-frequency power based on the fault-related cutoff frequency, and calculate the low-frequency power ratio sequence.

[0103] The relative modulation sequence is as follows:

[0104] (10);

[0105] in, For fluctuation components, As a trend component, The regularization constant, used to avoid numerical instability caused by excessively small local baselines, is set to 0.001 in this specific embodiment.

[0106] This relative modulation sequence separates low-frequency dynamic disturbances from the total envelope amplitude through a “local fluctuation / local baseline” approach.

[0107] Before power spectrum estimation, the relative modulation sequence is downsampled to prevent aliasing, resulting in the downsampled relative modulation sequence. and in the The power spectral density is estimated using the Welch method within each analysis window. ,in, For discrete frequency sampling point index, The highest frequency sampling point number corresponding to the upper limit of the analysis frequency;

[0108] The fault-related low-frequency power is:

[0109] (11);

[0110] The fault-related high-frequency power is:

[0111] (12);

[0112] in, This refers to the frequency sampling point number corresponding to the low-frequency cutoff frequency. Frequency resolution;

[0113] The low-frequency power ratio sequence is then:

[0114] (13);

[0115] in, is the power regularization constant.

[0116] When the degree of oil ingress fault in the servo joint increases, the low-frequency correlation energy of the current envelope relative to the modulation sequence increases accordingly, and the low-frequency power ratio shows an upward trend.

[0117] S7: The original current features obtained in step S2 and the low-frequency power ratio sequence obtained in step S6 are aligned in time length, normalized in scale, and fused to construct the original current-low-frequency power ratio fused health characterization.

[0118] Among them, let the first The original current characteristics of each period sample are as follows The low-frequency power ratio sequence is Scale normalization is performed separately:

[0119] (14);

[0120] (15);

[0121] in , , and The mean and standard deviation are obtained from the training samples. , It is a numerically stable term;

[0122] The original current-low-frequency power ratio combined health characterization is as follows:

[0123] (16);

[0124] in This is for sequential splicing.

[0125] The original current-low frequency power ratio fusion health characterization retains the local waveform details, transient changes and boundary transition information in the original current on the one hand, and introduces the fault-related low frequency modulation information in the low frequency power ratio on the other hand, thus forming a health characterization quantity that is more sensitive to the degree of oil inlet fault.

[0126] S8: Based on the original current-low frequency power ratio fusion health characterization, perform health status identification and output the health status identification result corresponding to the severity of the oil inlet fault.

[0127] The health status identification result includes at least one of the following: normal status, minor fault, moderate fault, serious fault, or serious fault.

[0128] The present invention will now be described in detail with reference to specific embodiments.

[0129] Current data preprocessing and raw current characterization construction:

[0130] In this embodiment, the input is the three-phase stator current timing data corresponding to the servo joints of the industrial robot. For continuous timing data, period alignment is first performed based on repetitive operation cycles, current boundaries, or label prediction results to obtain periodic samples with clear boundaries and uniform length. Subsequently, the periodic samples are scaled to form the original current characterization. This characterization is used to preserve local waveform morphology, transient changes, boundary transitions, and fine-grained timing textures.

[0131] Low-frequency power ratio sequence extraction:

[0132] like Figure 2 As shown, for each periodic sample, Clarke transform and Park transform are performed to construct a complex vector in the current space. Bandpass filtering is used to suppress high-frequency switching noise and broadband disturbances, while retaining narrowband signals relevant to servo joint operation. Let the narrowband current signal be... Constructing analytic signals And obtain the amplitude envelope Discretization yields the envelope sequence. .right Perform trend-volume decomposition to obtain trend components. and fluctuation components Further construct the relative modulation sequence The Welch method was used to estimate the power spectral density and calculate the low-frequency power ratio sequence. .

[0133] As the severity of the oil inlet fault increases, the low-frequency correlation energy of the current envelope relative to the modulation sequence gradually strengthens, and the low-frequency power ratio index value rises accordingly. The low-frequency power ratio sequence can highlight this type of low-frequency envelope modulation change, providing enhanced mechanistic information for fusion health characterization.

[0134] Raw current – ​​low-frequency power ratio combined health characterization:

[0135] Let the original current be characterized as The low-frequency power ratio sequence is First, unify both to the same time length or analysis window index, and then perform scale normalization separately to obtain... and Subsequently, a fused health characterization was constructed using a method of first branching modeling and then adaptive fusion. The dataset corresponding to this fused health representation. - ,in Represents the original current information. This indicates the low-frequency power ratio information extracted from the original current.

[0136] The core function of this fusion method lies in utilizing two types of complementary information: the original current retains local waveform morphology, transient changes, and fine-grained texture, while the low-frequency power ratio enhances trend terms and low-frequency fluctuation information consistent with the oil inlet fault mechanism. Combining these two methods improves the stability of distinguishing adjacent fault levels.

[0137] Health monitoring output and algorithm validation results:

[0138] During the health monitoring phase, health status is identified based on fused health representations, and the category result corresponding to normal status or failure is output, such as... Figure 3 As shown, a is the current waveform of a normal sample, b is the current waveform of a fault sample, c is the low-frequency power ratio characteristic of a normal sample, and d is the low-frequency power ratio characteristic of a fault sample.

[0139] To demonstrate the effectiveness of the method of this invention, the original current datasets were compared under the same model and evaluation settings. Low-frequency power ratio dataset and fused datasets - The health status recognition effect.

[0140] The verification results show that the fused dataset can simultaneously utilize the local details of the original current and the low-frequency power ratio mechanism to enhance information, and its recognition performance is better than that of a single input.

[0141] Table 1

[0142]

[0143] Table 1 shows a comparison of algorithm evaluation metrics for different input representations. The results are only used to illustrate the effectiveness of the method flow of this invention: When using the low-frequency power ratio alone, although it can reflect the fault-related low-frequency modulation information, it will lose some local waveform details; when using the original current alone, although it retains local details, it is insufficient to explicitly express the low-frequency modulation law; after the two are fused, the confusion between adjacent fault levels can be reduced and the stability of health status identification can be improved.

[0144] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A method for identifying the health status of a joint motor with oil ingress faults by integrating current timing and low-frequency power ratio, characterized in that, Includes the following steps: S1: Obtain the three-phase stator current timing data of the servo joint motor to be monitored, and perform period alignment, sample segmentation and length unification on the three-phase stator current timing data to obtain the original current sample sequence; S2: Scale normalize the original current sample sequence and extract the original current features used to characterize local waveform morphology, transient changes and fine-grained temporal texture; S3: Perform Clarke and Park transformations on the three-phase stator current timing data to obtain the orthogonal current components in the two-phase stationary coordinate system and the current components in the synchronous rotating coordinate system, and construct a complex vector of current space based on the orthogonal current components. S4: Narrowband filtering is performed using the main conduction angular frequency of the current space complex vector as the center frequency to obtain the narrowband current signal in the fundamental frequency neighborhood; S5: Perform Hilbert envelope extraction on the narrowband current signal to obtain the amplitude envelope sequence, and perform trend-fluctuation decomposition on the amplitude envelope sequence to extract the trend component and fluctuation component; S6: Construct a relative modulation sequence based on the trend component and the fluctuation component, calculate the power spectrum of the relative modulation sequence, calculate the low-frequency power and high-frequency power based on the fault-related cutoff frequency, and calculate the low-frequency power ratio sequence. S7: The original current features obtained in step S2 and the low-frequency power ratio sequence obtained in step S6 are aligned in time length, normalized in scale, and fused to construct the original current-low-frequency power ratio fused health characterization. S8: Based on the original current-low frequency power ratio fusion health characterization, perform health status identification and output the health status identification result corresponding to the severity of the oil inlet fault.

2. The method for identifying the health status of a joint motor with oil inlet fault based on the fusion of current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S1, the period alignment includes dividing the continuous three-phase stator current time series data into period samples of uniform length according to the working cycle, so that samples of different health states are comparable in terms of time scale and sample length.

3. The method for identifying the health status of a joint motor with oil inlet fault based on the fusion of current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S3, the three-phase stator current timing data is processed. , and Perform Clarke transform to obtain Orthogonal current components of the axis and Orthogonal current components of the axis : (1); The complex vector of current space is calculated. : (2); in, It is the symbol for the imaginary unit in complex numbers.

4. The method for identifying the health status of a joint motor with oil inlet fault based on the fusion of current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S4, the bandpass filter cutoff frequency of the narrowband filter. and upper cutoff frequency Determine using the following method: First, define the lower bound for removing DC and slow-varying trends. And the upper limit for retaining the main modulation component of the current. ; Further order: (3); (4); in To ensure the upper cutoff frequency Greater than the lower cutoff frequency Positive numbers, This is below the safety limit for Nyquist frequencies.

5. The method for identifying the health status of a joint motor with oil inlet fault based on fusion current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S5, the amplitude envelope sequence is obtained from the analytic signal. for: (5); in, It is a narrowband current signal. For Hilbert transform, Amplitude envelope for: (6)。 6. The method for identifying the health status of a joint motor with oil inlet fault based on fusion current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S5, the trend component Obtained from the moving average operator: (7); in, It is a discrete amplitude envelope sequence. This represents the value of the discrete amplitude envelope sequence at the (n-q)th sampling point. The total number of samples in the discrete amplitude envelope sequence; The fluctuation component The calculation is as follows: (8); (9); in, It is a discrete amplitude envelope sequence. , T represents the fluctuation component of the discrete amplitude envelope sequence at the nth sampling point. s The sampling interval is... The length of the sliding window. The cutoff frequency is The low-pass filter operator.

7. The method for identifying the health status of a joint motor with oil inlet fault based on fusion current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S6, the relative modulation sequence is: (10); in, For fluctuation components, As a trend component, This is a regularization constant used to avoid numerical instability caused by excessively small local baselines.

8. The method for identifying the health status of a joint motor with oil inlet fault based on the fusion of current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S6, the relative modulation sequence is downsampled to prevent aliasing before power spectrum estimation to obtain the downsampled relative modulation sequence. and in the The power spectral density is estimated using the Welch method within each analysis window. ,in, For discrete frequency sampling point index, The highest frequency sampling point number corresponding to the upper limit of the analysis frequency; The fault-related low-frequency power is: (11); The fault-related high-frequency power is: (12); Where, is the low-frequency cutoff frequency. This refers to the frequency sampling point number corresponding to the low-frequency cutoff frequency. Frequency resolution; The low-frequency power ratio sequence is then: (13); in, is the power regularization constant.

9. The method for identifying the health status of a joint motor with oil inlet fault based on the fusion of current timing and low-frequency power ratio as described in claim 1, characterized in that, In step S7, let the first... The original current characteristics of each period sample are as follows The low-frequency power ratio sequence is Scale normalization is performed separately: (14); (15); in , , and The mean and standard deviation are obtained from the training samples. , It is a numerically stable term; The original current-low-frequency power ratio combined health characterization is as follows: (16); in This is for sequential splicing.

10. The method for identifying the health status of a joint motor with oil inlet fault based on fusion current timing and low-frequency power ratio according to claim 1, characterized in that, In step S8, the health status identification result includes at least one of normal status, minor fault, secondary serious fault, more serious fault, or serious fault.