DD motor residual life prediction method based on multi-source data fusion

Abstract

The invention relates to the technical field of data processing, in particular to a DD motor residual life prediction method based on multi-source data fusion, and the method comprises the steps: obtaining a vibration signal, a q-axis torque current and a rotating speed signal of a current time window of a DD motor; constructing frequency spectrum complexity based on the vibration frequency domain distribution characteristics and the q-axis current load fluctuation characteristics; constructing a non-stability constraint coefficient based on the rotating speed transient characteristic and the electromechanical coupling power characteristic; calculating an optimal modal decomposition layer number and an optimal secondary penalty factor of variational modal decomposition through nonlinear mapping; decomposing the vibration signal, taking a kurtosis maximum intrinsic mode function component as a main fault component, and extracting a health factor; and predicting the remaining life based on the health factor change trend. The DD motor residual life prediction precision and reliability are improved, fault features are accurately extracted, under-decomposition or over-decomposition is avoided, and a scientific basis is provided for preventive maintenance.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method for predicting the remaining life of a DD motor based on multi-source data fusion. Background Technology

[0002] Direct Drive Motors (DD Motors) are used in high-end manufacturing fields such as semiconductor wafer transport and precision CNC machine tools due to their characteristics of no reduction gear, zero backlash, and high rigidity. In practical applications, in order to meet the needs of high-efficiency production, DD Motors usually need to perform frequent start-stop, speed change, and variable load cutting tasks. Due to the lack of a reduction gear to buffer the impact, the impact at the load end will be directly fed back to the motor body, resulting in an extremely complex working condition for DD Motors. In order to ensure production quality and avoid unplanned downtime, it is particularly important to predict the remaining life of key components of DD Motors, such as bearings and stator windings.

[0003] Currently, Variational Mode Decomposition (VMD), as an adaptive signal processing method, is often used to extract fault features of rotating machinery due to its advantages in suppressing mode aliasing and noise resistance. VMD decomposes the signal into several intrinsic mode functions (IMFs) with specific center frequencies by iteratively searching for the extrema of the variational model.

[0004] However, when existing VMD-based fault diagnosis methods are applied to DD motors, the effectiveness of VMD is highly dependent on pre-set key parameters, namely the number of mode decomposition layers and the second-order penalty factor. Existing technologies usually set the key parameters to fixed values ​​based on experience, or determine them only through a single optimization for a certain stable operating condition. However, in actual operation, the speed and load torque of DD motors change drastically in real time with the manufacturing process, exhibiting typical variable operating condition characteristics. This results in the dynamic time-varying characteristics of the spectral structure and bandwidth of the vibration signal. If fixed parameters are used, when the operating condition becomes complex, such as during multi-order resonance excitation, if the fixed number of mode decomposition layers is too small, it will lead to under-decomposition, causing weak fault features to be submerged in background noise or strong interference. When the operating condition changes rapidly, such as during rapid acceleration or deceleration, a fixed second-order penalty factor may result in an excessively narrow filtering bandwidth, thus losing broadband frequency modulation fault information caused by speed fluctuations, or introducing too much noise due to an excessively wide bandwidth, ultimately affecting the accuracy of the remaining life prediction of the DD motor. Summary of the Invention

[0005] To address the problem that traditional VMD (Virtual Machine Determination) with its fixed parameters struggles to effectively extract subtle early-stage fault features of DD (Displacement Damped) motors, leading to distorted health factors and ultimately affecting the accuracy of DD motor remaining life prediction, this invention proposes a DD motor remaining life prediction method based on multi-source data fusion. This method includes the following steps: Multi-source data of the DD motor within the current time window is acquired, including vibration signals, q-axis torque current, and speed signals. Based on the frequency domain distribution characteristics of the vibration signals and the load fluctuation characteristics of the q-axis torque current signals, a spectral complexity reflecting the degree of disorder in the signal frequency distribution within the current time window is constructed. Based on the transient characteristics and electromechanical coupling power characteristics of the speed signals, a non-stationarity constraint coefficient reflecting the degree of non-stationarity of the signals within the current time window is constructed. Based on the spectral complexity and the non-stationarity constraint coefficient, the optimal mode decomposition level and the optimal quadratic penalty factor of the variational mode decomposition at the current moment are calculated through a nonlinear mapping relationship. Using the optimal mode decomposition level and the optimal quadratic penalty factor, variational mode decomposition is performed on the vibration signals to obtain several intrinsic mode function components. The intrinsic mode function component with the largest kurtosis is taken as the main fault component, and a health factor of the main fault component is extracted. Based on the changing trend of the health factor, the remaining life of the DD motor is predicted.

[0006] The beneficial effects are as follows: By constructing a calculation model of spectral complexity and non-stationarity constraint coefficients, a comprehensive evaluation of the operating conditions of DD motors is achieved, accurately reflecting the dynamic characteristics of signals under varying operating conditions, and providing a reliable operating condition perception basis for adaptive adjustment of VMD parameters; by calculating the optimal VMD parameters through nonlinear mapping relationships, adaptive optimization of the mode decomposition process is achieved, ensuring accurate extraction of fault features under complex varying operating conditions and avoiding under-decomposition or over-decomposition problems caused by fixed parameters; by extracting the intrinsic mode function with the largest kurtosis as the main fault component, accurate identification of weak fault features is achieved, effectively improving the accuracy of fault feature extraction and providing high-quality health status indicators for remaining lifetime prediction; and remaining lifetime prediction based on the changing trend of health factors improves prediction accuracy and reliability, providing a scientific basis for preventive maintenance of DD motors.

[0007] Furthermore, the spectral complexity satisfies: In the formula, The spectral complexity is within the current time window. The power spectral density of the vibration signal within the current time window at frequency The value at that location, This represents the total number of effective frequency points of the power spectral density of the vibration signal within the current time window. The load weighting factor is set. The time within the current time window q-axis torque current, This represents the average value of the q-axis torque current within the current time window. This is the rated operating current of the DD motor. This represents the total number of moments within the current time window. It is the natural logarithm function.

[0008] The beneficial effects are as follows: by constructing a composite function that includes power spectral entropy and load fluctuation variance, a comprehensive evaluation of spectral complexity is achieved. The power spectral entropy term accurately reflects the degree of disorder in the frequency domain distribution of the vibration signal, and the load fluctuation variance term reflects the fluctuation characteristics of the q-axis torque current. Thus, the spectral complexity of the DD motor under varying operating conditions can be accurately evaluated, providing a reliable frequency domain characteristic basis for VMD parameter optimization.

[0009] Furthermore, the load weighting coefficient is determined through training with historical fault data, and its value range is [range missing]. .

[0010] Furthermore, the nonstationarity constraint coefficients satisfy: In the formula, The coefficients representing the non-stationarity constraints within the current time window. The time within the current time window The absolute value of the rate of change of rotational speed. The time within the current time window The differential increment of the rotational speed signal, For time increments, This represents the total number of moments within the current time window. The time within the current time window q-axis torque current, The time within the current time window The rotational speed signal, This refers to the rated output power of the DD motor. The length of the current time window. The integral symbol is used. This is a normalized function based on the rated angular acceleration of the DD motor. It is a function with maximum value. It is a natural exponential function.

[0011] The beneficial effects are as follows: by constructing a product term that includes the maximum speed change rate and the electromechanical coupling power integral, a scientific evaluation of the non-stationarity constraint coefficient is achieved. The maximum speed change rate reflects the transient characteristics of the speed, and the electromechanical coupling power integral reflects the coupling effect between current and speed, thereby accurately evaluating the non-stationarity of the DD motor and providing a reliable time-domain characteristic basis for the adaptive adjustment of VMD parameters.

[0012] Furthermore, the optimal number of mode decomposition layers satisfies: In the formula, The optimal mode decomposition level at the current moment. The number of fundamental modes in variational mode decomposition. This is the first adjustment parameter used to control the degree to which spectral complexity affects the optimal number of mode decomposition layers. The spectral complexity is within the current time window. It is a function with maximum value. This is the rounding function. It is the natural logarithm function.

[0013] The beneficial effects are as follows: by constructing a linear combination containing the number of basic modes and the logarithm of spectral complexity, the adaptive calculation of the optimal number of mode decomposition layers is realized, ensuring that the number of mode decomposition layers can be dynamically adjusted according to the signal complexity. When the spectral complexity is high, the number of decomposition layers is increased to fully extract fault features, and when the complexity is low, the number of decomposition layers is reduced to avoid over-decomposition, thereby improving the adaptability of VMD under varying operating conditions.

[0014] Furthermore, the number of basic modes is determined based on typical fault characteristics of DD motors, and its value range is [range missing]. .

[0015] Furthermore, the optimal quadratic penalty factor satisfies: In the formula, The optimal quadratic penalty factor at the current moment. The maximum bandwidth constraint benchmark value for the quadratic penalty factor in variational mode decomposition is given. This is the second adjustment parameter used to control the degree of influence of the non-stationarity constraint coefficient on the optimal quadratic penalty factor. This represents the non-stationarity constraint coefficient within the current time window.

[0016] The beneficial effects are as follows: by constructing a fractional function containing the maximum bandwidth constraint and non-stationarity constraint coefficients, the optimal quadratic penalty factor is adaptively calculated, ensuring that the penalty factor can be dynamically adjusted according to the non-stationarity of the signal. When the non-stationarity is high, the penalty factor is reduced to adapt to the broadband frequency modulation characteristics, and when the non-stationarity is low, the penalty factor is increased to enhance the noise resistance, thereby improving the filtering performance of VMD under varying operating conditions.

[0017] Furthermore, the maximum bandwidth constraint benchmark value is the recommended maximum bandwidth constraint value under stable operating conditions, and its value range is [range missing]. .

[0018] Further, the extraction of the health factor of the main fault component includes: performing a Hilbert transform on the main fault component to obtain the envelope signal of the main fault component; calculating the effective value of the envelope signal as a first health factor; and calculating the energy entropy of the envelope signal as a second health factor.

[0019] Furthermore, the prediction of the remaining lifespan of the DD motor includes: constructing a degradation model based on the Wiener process, inputting the health factors of the main fault components corresponding to historical moments and the current moment into the degradation model and updating the model parameters; predicting the remaining time required for the health factors to reach a set failure threshold, and using this remaining time as the remaining lifespan of the DD motor.

[0020] The present invention has the following beneficial effects: (1) In view of the problem that the fixed mode decomposition layer and the second-order penalty factor of traditional VMD cannot adapt to the dynamic changes in signal spectrum complexity and non-stationarity caused by frequent start-stop and variable speed and load of DD motor, this invention constructs the spectrum complexity by using the frequency domain distribution characteristics of vibration signal and the q-axis torque current load fluctuation characteristics, and constructs the non-stationarity constraint coefficient by combining the speed transient characteristics and electromechanical coupling power characteristics. The optimal parameters are dynamically calculated through nonlinear mapping relationship. When the working conditions are complex, such as multi-order resonance, the optimal mode decomposition layer automatically increases to avoid under-decomposition and the submergence of weak fault features. When the working conditions are transient, such as rapid acceleration and deceleration, the optimal second-order penalty factor is adaptively adjusted to ensure that the filter frequency band matches the signal bandwidth characteristics, so as not to lose broadband frequency modulation fault information or introduce too much noise. This effectively solves the problem of decomposition failure of fixed parameters under changing working conditions and lays an accurate foundation for fault feature extraction.

[0021] (2) Breaking through the limitations of traditional VMD which relies solely on a single vibration signal, this method integrates multi-source information, including vibration signals reflecting the dynamic characteristics of mechanical faults, q-axis torque and current reflecting the characteristics of load impact and electromagnetic coupling, and speed signals reflecting transient operating conditions. This allows the calculation of spectral complexity and non-stationarity constraint coefficients to fully characterize the operating conditions and signal essence of the DD motor. Compared to a single signal, multi-source data can complement and cancel out the noise interference of a single data source, more accurately capturing the coupling relationship between load impact, speed change, and signal characteristics. This makes the calculation of optimal parameters more consistent with actual operating conditions and provides richer feature support for the subsequent screening of main fault components, avoiding feature extraction deviations caused by the one-sidedness of single data.

[0022] (3) By selecting the intrinsic mode function with the largest kurtosis as the main fault component, the high sensitivity of kurtosis to impact faults, such as bearing wear and stator winding partial discharge, is fully utilized to accurately remove background noise and non-fault interference components in the vibration signal, such as mechanical vibration and electromagnetic interference under normal working conditions, so that the main fault component can reflect the deterioration state of key components of the DD motor. Based on the health factor extracted from the main fault component, the correlation between weak faults, component deterioration and life decay can be effectively amplified, avoiding interference of non-fault characteristics on health status assessment, ensuring that the changing trend of health factor can truly and sensitively reflect the remaining life status of the DD motor, and improving the accuracy of life prediction. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the steps of a method for predicting the remaining life of a DD motor based on multi-source data fusion, according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the health factor degradation trend of a DD motor remaining life prediction method based on multi-source data fusion according to an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below. The described embodiments are only a part of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] Please see Figure 1 The diagram illustrates a flowchart of a method for predicting the remaining life of a DD motor based on multi-source data fusion, according to an embodiment of the present invention. The method includes the following steps: S001: Obtain multi-source data of the DD motor within the current time window, including vibration signal, q-axis torque current and speed signal.

[0028] Specifically, vibration signals are collected by an accelerometer installed on the DD motor body, and stator q-axis torque, current and speed signals are collected synchronously through the communication interface of the servo driver. In order to eliminate signal drift and timing deviation, all collected signals are detrended and timestamp aligned, and a preset current time window, such as a multi-source data segment within 5 seconds, is extracted. For example, the acquisition frequency is 10kHz, which is recorded as the multi-source data of the DD motor in the current time window.

[0029] S002: Based on the frequency domain distribution characteristics of the vibration signal and the load fluctuation characteristics of the q-axis torque current signal, construct a spectral complexity that reflects the degree of disorder in the signal frequency distribution within the current time window.

[0030] It should be noted that DD motors, used in precision CNC machine tool cutting and semiconductor wafer transport, frequently endure varying load impacts and speed changes. The frequency domain distribution of their vibration signals exhibits dynamic discreteness due to multi-order resonance excitation and the superposition of load disturbances. The frequency domain characteristics of the vibration signal alone cannot fully reflect the coupling effect of frequency disorder and load fluctuations. Furthermore, the q-axis torque current is directly related to the motor's output torque, and its fluctuation amplitude is positively correlated with the load impact intensity, effectively characterizing the disturbance effect of operating conditions on the signal spectrum. Therefore, it is necessary to integrate both to construct a comprehensive index. Thus, this step combines the power spectral entropy of the vibration signal with the normalized variance of the q-axis torque current to construct spectral complexity, achieving accurate quantification of the degree of frequency distribution disorder under varying operating conditions. This provides a physical basis for the adaptive adjustment of the number of VMD mode decomposition layers in the subsequent process.

[0031] Specifically, the spectral complexity satisfies: ; In the formula, The spectral complexity within the current time window is dimensionless. The power spectral density of the vibration signal within the current time window at frequency The value at that location is calculated after performing an FFT transform on the vibration signal, and the unit is . ; The total number of effective frequency points of the power spectral density of the vibration signal within the current time window, determined based on the sampling frequency and signal characteristics, is dimensionless. The load weighting factor is set, for example, the range of values ​​is: It is determined through training with historical fault data and is dimensionless; The time within the current time window The q-axis torque current, in units of ; This represents the average value of the q-axis torque current within the current time window. This refers to the rated operating current of the DD motor, in units of... ; This represents the total number of moments within the current time window. It is the natural logarithm function.

[0032] in, The information entropy of the power spectrum is as follows: when the DD motor is in complex cutting or high-speed operation and excites multi-order resonance, the spectrum energy distribution becomes discrete and the information entropy increases. The normalized torque-current variance reflects the severity of load fluctuations. Therefore, the spectral complexity is positively correlated with the richness of frequency components in the scene and the degree of load disturbance, providing a physical basis for subsequent determination of the number of modes.

[0033] S003: Based on the transient characteristics and electromechanical coupling power characteristics of the speed signal, construct a non-stationarity constraint coefficient that reflects the degree of non-stationarity of the signal within the current time window.

[0034] It should be noted that under transient conditions such as rapid acceleration / deceleration and heavy-load start-stop, the non-stationarity of the DD motor signal mainly stems from two aspects: first, the imbalance of motion caused by sudden changes in speed; and second, the electromechanical characteristic fluctuations caused by the coupling between load and speed. The rate of change of speed alone cannot cover the enhanced non-stationarity effect brought about by the coupling between load and speed. However, the electromechanical coupling power can directly reflect the synergistic effect of the two, and its integral value can quantify the cumulative degree of non-stationarity over a period of time. Therefore, this step integrates the transient speed characteristics and the electromechanical coupling power characteristics to construct a non-stationarity constraint coefficient, more accurately characterizing the degree of signal non-stationarity under varying operating conditions, and providing a time-domain basis for the subsequent adaptive adjustment of the VMD secondary penalty factor.

[0035] Specifically, the nonstationarity constraint coefficients satisfy: ; In the formula, This represents the non-stationarity constraint coefficient within the current time window; it is dimensionless. The time within the current time window The absolute value of the rate of change of rotational speed. The time within the current time window The differential increment of the rotational speed signal, For time increments, This represents the total number of moments within the current time window. The time within the current time window q-axis torque current, The time within the current time window The rotational speed signal, This refers to the rated output power of the DD motor. The length of the current time window. The integral symbol is used. This is a normalized function based on the rated angular acceleration of the DD motor, and its value is the input quantity divided by the rated angular acceleration of the DD motor. It is a function with maximum value. It is a natural exponential function.

[0036] in, It directly characterizes the transient nature of motion states; The ratio of average load power to rated power represents the load level. When the DD motor accelerates or decelerates rapidly or operates under heavy load with high power, the non-stationarity and frequency modulation of the signal are significantly enhanced. By utilizing the nonlinear amplification characteristics of the exponential function, the non-stationarity constraint coefficient can keenly capture the fault diffusion trend under high load, thereby reflecting the degree of demand for wideband decomposition in the scenario. The larger the non-stationarity constraint coefficient, the more the bandwidth limit must be relaxed.

[0037] S004: Based on the spectral complexity and the non-stationarity constraint coefficient, calculate the optimal mode decomposition level and the optimal quadratic penalty factor of the variational mode decomposition at the current moment through a nonlinear mapping relationship.

[0038] It should be noted that the number of mode decomposition layers in VMD determines the precision of signal decomposition, while the second-order penalty factor determines the filtering bandwidth during decomposition. When the spectral complexity is high, the number of mode layers needs to be increased to avoid under-decomposition; when the non-stationarity is high, the second-order penalty factor needs to be reduced to broaden the filtering bandwidth and avoid losing broadband frequency modulation fault information. Traditional fixed parameters cannot adapt to the dynamic changes in the variable operating conditions of DD motors. Therefore, a nonlinear mapping relationship needs to be established between spectral complexity and the number of mode layers, and between the non-stationarity constraint coefficient and the second-order penalty factor, to achieve dynamic parameter matching. Thus, this step constructs a mapping relationship by progressively adjusting the number of mode layers and inversely adjusting the second-order penalty factor to calculate the optimal VMD parameters that precisely match the current operating conditions, thereby solving the decomposition failure problem caused by fixed parameters.

[0039] Specifically, the optimal number of mode decomposition layers satisfies: ; In the formula, The optimal mode decomposition level at the current moment. The fundamental mode number is the number of variational mode decomposition modes. This number is determined based on typical fault characteristics of the DD motor, and its value range is [range missing]. ; The first adjustment parameter, used to control the influence of spectral complexity on the optimal number of mode decomposition layers, is determined through training and optimization using historical data, and its value range is [value range missing]. , The spectral complexity is within the current time window. It is a function with maximum value. This is the rounding function. It is the natural logarithm function.

[0040] Among them, as the spectral complexity increases, the number of optimal mode decomposition layers increases logarithmically, thereby adaptively increasing the number of decomposed modes and preventing feature aliasing caused by under-decomposition.

[0041] Specifically, the optimal quadratic penalty factor satisfies: ; In the formula, The optimal quadratic penalty factor at the current moment. The maximum bandwidth constraint benchmark value is the second-order penalty factor for variational mode decomposition. This benchmark value is the recommended maximum bandwidth constraint value under stationary conditions, and its range is [value missing]. ; The second adjustment parameter, used to control the influence of the non-stationarity constraint coefficient on the optimal quadratic penalty factor, is determined through training and optimization using historical data, and its value range is [value range missing]. ; This represents the non-stationarity constraint coefficient within the current time window.

[0042] As the nonstationarity constraint coefficient increases, the denominator increases, leading to a decrease in the optimal quadratic penalty factor. According to the VMD principle, a smaller optimal quadratic penalty factor corresponds to a wider bandwidth, thereby ensuring that the broadband modulation signal caused by the fault can be completely preserved in the same IMF component under nonstationary conditions.

[0043] S005: Using the optimal number of mode decomposition layers and the optimal quadratic penalty factor, perform variational mode decomposition on the vibration signal to obtain several intrinsic mode function components. Take the intrinsic mode function component with the largest kurtosis as the main fault component and extract the health factor of the main fault component.

[0044] It should be noted that the multiple IMF components obtained from VMD decomposition include non-fault components such as background noise, normal mechanical vibration, and electromagnetic interference, as well as fault-induced characteristic components such as bearing wear and stator winding partial discharge. Fault characteristics, especially impact faults, will cause the corresponding IMF components to exhibit significant impact characteristics. Kurtosis, as a sensitive indicator of signal impact, indicates that the larger the value, the more prominent the fault impact characteristics contained in the IMF component. Therefore, this step, by screening the IMF component with the largest kurtosis, can accurately isolate non-fault interference and noise, and locate the main fault component. Then, the envelope signal is extracted through Hilbert transform to highlight the amplitude changes of fault characteristics, and the effective value and energy entropy are calculated as health factors to achieve the assessment of the motor's health status.

[0045] Specifically, the health factors for extracting the main fault component include: The main fault component is subjected to Hilbert transform to obtain the analytic signal, and the amplitude sequence of the analytic signal is extracted as the envelope signal of the main fault component. The effective value of the envelope signal is calculated as the first health factor; The energy entropy of the envelope signal is calculated as a second health factor.

[0046] S006: Based on the changing trends of the aforementioned health factors, predict the remaining lifespan of the DD motor.

[0047] It is important to note that the remaining lifespan of a DD motor is essentially the degradation process of its key components, such as bearings and stator windings, from their current healthy state to a failure state. Health factors evolve systematically with component deterioration, such as a gradual increase in effective values ​​and energy entropy. The Wiener process has the advantage of characterizing gradual degradation and random fluctuations, effectively adapting to the dynamic changes in health factors. Furthermore, it can continuously update model parameters using historical and current data, improving the real-time performance and accuracy of predictions. Therefore, this step constructs a degradation model using the Wiener process to dynamically track the changing trends of health factors. Combined with failure thresholds determined based on historical fault data, such as the critical value of the effective envelope signal corresponding to bearing wear and the critical value of energy entropy corresponding to stator winding aging, this more accurately predicts the remaining time required for health factors to reach the threshold, providing a precise time reference for preventative maintenance of DD motors.

[0048] Specifically, the prediction of the remaining lifespan of the DD motor includes: A degradation model is constructed based on the Wiener process. The health factors of the main fault components corresponding to the historical time and the current time are input into the degradation model and the model parameters are updated. The remaining time required for the health factor to reach the set failure threshold is predicted. For example, the failure threshold corresponding to the first health factor is 3 and the failure threshold corresponding to the second health factor is 8. The minimum of the remaining time of the two health factors is taken as the remaining life of the DD motor.

[0049] like Figure 2 As shown, this diagram visually illustrates the evolution of health factors and the failure threshold determination logic during DD motor operation. The horizontal axis represents the DD motor's operating time, and the unit can be set to hours, operating cycles, etc., depending on the actual application scenario. In this embodiment, the unit is a dimensionless operating cycle, with a value range of... The vertical axis represents the health factor value; curve H1 corresponds to the first health factor and curve H2 corresponds to the second health factor; the two horizontal dashed lines are the failure thresholds of H1 and H2, respectively; as can be seen from the simulation trend, as the DD motor runs for longer, key components, such as bearings and stator windings, gradually deteriorate, and both H1 and H2 show a monotonically increasing trend: the increase of H1 reflects the continuous enhancement of the amplitude of the fault impact signal, and the increase of H2 reflects the increasingly discrete frequency domain energy distribution corresponding to the fault characteristics. When any health factor reaches or exceeds its corresponding failure threshold, the key component of the DD motor is determined to have entered a failure state, and the time difference between the current moment and the failure moment is the predicted remaining life.

[0050] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the remaining life of a DD motor based on multi-source data fusion, characterized in that, include: Acquire multi-source data of the DD motor within the current time window, including vibration signal, q-axis torque current and speed signal; Based on the frequency domain distribution characteristics of the vibration signal and the load fluctuation characteristics of the q-axis torque current signal, a spectral complexity reflecting the degree of disorder in the signal frequency distribution within the current time window is constructed. Based on the transient characteristics and electromechanical coupling power characteristics of the speed signal, a non-stationarity constraint coefficient reflecting the degree of non-stationarity of the signal within the current time window is constructed. Based on the spectral complexity and the non-stationarity constraint coefficient, the optimal mode decomposition level and the optimal quadratic penalty factor of the variational mode decomposition at the current moment are calculated through a nonlinear mapping relationship. Using the optimal number of mode decomposition layers and the optimal quadratic penalty factor, variational mode decomposition is performed on the vibration signal to obtain several intrinsic mode function components. The intrinsic mode function component with the largest kurtosis is taken as the main fault component, and the health factor of the main fault component is extracted. Based on the changing trends of the aforementioned health factors, the remaining lifespan of the DD motor is predicted.

2. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 1, characterized in that, The spectral complexity satisfies: ； In the formula, The spectral complexity is within the current time window. The power spectral density of the vibration signal within the current time window at frequency The value at that location, This represents the total number of effective frequency points of the power spectral density of the vibration signal within the current time window. The load weighting factor is set. The time within the current time window q-axis torque current, This represents the average value of the q-axis torque current within the current time window. This is the rated operating current of the DD motor. This represents the total number of moments within the current time window. It is the natural logarithm function.

3. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 2, characterized in that, The load weighting coefficient is determined through training on historical fault data, and its value range is [value range missing]. .

4. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 1, characterized in that, The nonstationarity constraint coefficients satisfy: ； In the formula, The coefficients representing the non-stationarity constraints within the current time window. The time within the current time window The absolute value of the rate of change of rotational speed. The time within the current time window The differential increment of the rotational speed signal, For time increments, This represents the total number of moments within the current time window. The time within the current time window q-axis torque current, The time within the current time window The rotational speed signal, This refers to the rated output power of the DD motor. The length of the current time window. The integral symbol is used. This is a normalized function based on the rated angular acceleration of the DD motor. It is a function with maximum value. It is a natural exponential function.

5. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 1, characterized in that, The optimal number of mode decomposition layers satisfies: ； In the formula, The optimal mode decomposition level at the current moment. The number of fundamental modes in variational mode decomposition. This is the first adjustment parameter used to control the degree to which spectral complexity affects the optimal number of mode decomposition layers. The spectral complexity is within the current time window. It is a function with maximum value. This is the rounding function. It is the natural logarithm function.

6. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 5, characterized in that, The fundamental mode number is determined based on typical fault characteristics of DD motors, and its value range is [value range missing]. .

7. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 1, characterized in that, The optimal quadratic penalty factor satisfies: ； In the formula, The optimal quadratic penalty factor at the current moment. The maximum bandwidth constraint benchmark value for the quadratic penalty factor in variational mode decomposition is given. This is the second adjustment parameter used to control the degree of influence of the non-stationarity constraint coefficient on the optimal quadratic penalty factor. This represents the non-stationarity constraint coefficient within the current time window.

8. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 7, characterized in that, The maximum bandwidth constraint benchmark value is the recommended maximum bandwidth constraint value under stable operating conditions, and its value range is [value missing]. .

9. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 1, characterized in that, The health factors for extracting the main fault component include: The envelope signal of the main fault component is obtained by performing a Hilbert transform on the main fault component. The effective value of the envelope signal is calculated as the first health factor; The energy entropy of the envelope signal is calculated as a second health factor.

10. The method for predicting the remaining life of a DD motor based on multi-source data fusion according to claim 1, characterized in that, The prediction of the remaining lifespan of the DD motor includes: A degradation model is constructed based on the Wiener process. The health factors of the main fault components corresponding to the historical time and the current time are input into the degradation model and the model parameters are updated. Predict the remaining time required for the health factor to reach a set failure threshold, and use this remaining time as the remaining lifespan of the DD motor.

Images