Motor stator fault detection system and detection method

Abstract

The invention provides a motor stator fault detection system and method, and relates to the technical field of data processing, and the method comprises the steps: synchronously collecting a three-phase current signal and a stator winding temperature signal during the operation of a motor in real time through a current sensor and a temperature sensor; performing time-frequency domain analysis and spectrum feature extraction on the three-phase current signal to obtain a current unbalance degree and a current harmonic distortion rate, and forming a current feature parameter set; performing filtering noise reduction and trend elimination processing on the stator winding temperature signal to obtain a processed signal; on the basis of the processed signals, temperature abnormal points are recognized through a k-nearest neighbor clustering algorithm of the three-dimensional Euclidean distance, the temperature instantaneous change rate is calculated according to the temperature abnormal points, and a temperature characteristic parameter set containing the abnormal point density and the temperature rise rate is obtained. According to the invention, stator fault detection and differential early warning are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a motor stator fault detection system and a detection method. BACKGROUND

[0002] In the field of industrial production, motors are widely used in automobile manufacturing, steel smelting, and power generation as core power equipment. The stator winding, as a key component of motor energy conversion, will directly cause motor downtime and even equipment burning and production interruption due to insulation aging and turn-to-turn short circuit faults. Therefore, real-time and accurate detection of stator faults is crucial to ensure the stable operation of industrial systems.

[0003] The current commonly used motor stator fault detection method in the industry relies on single parameter monitoring. For example, a servo motor detection system for a new energy automobile battery production line uses a traditional current transformer to collect real-time three-phase current signals to calculate the current imbalance degree, and collects temperature data through a fixed-point temperature sensor pasted on the surface of the stator winding. The system triggers an early warning only when the current imbalance degree exceeds the preset threshold or the temperature reaches the warning value. However, in actual production, the servo motor needs to be frequently started and stopped and the load needs to be frequently switched due to the battery assembly process requirements. At this time, the current signal is distorted due to the sudden change of the load, and the system mistakenly judges the normal load change as a current imbalance fault, triggering 12 invalid early warnings in a single month, interfering with normal production scheduling. When a local micro-short circuit occurs in the stator winding of a servo motor due to insulation material aging, the fixed-point temperature sensor is disturbed by the air conditioning airflow in the workshop and fails to capture the local temperature rise anomaly inside the winding, resulting in the fault not being detected in time, and eventually causing the stator winding to burn out, causing the production line to shut down. The technical defects of the traditional detection method rely only on single current or temperature parameter monitoring, lack multi-parameter collaborative analysis capability, and are difficult to distinguish between normal operating fluctuations and fault signals. The anti-environmental interference ability is weak, the sensitivity to local faults is insufficient, and the dynamic influence of operating conditions on the detection threshold is not considered, resulting in high false alarm rate and high risk of missed reports. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a motor stator fault detection system and a detection method to realize the detection and differentiated early warning of stator faults.

[0005] To solve the above technical problems, the technical solution of the present application is as follows: In a first aspect, a motor stator fault detection method is provided, the method comprising: real-time and synchronous acquisition of three-phase current signals and stator winding temperature signals during motor operation through current sensors and temperature sensors; The three-phase current signal is analyzed in time-frequency domain and spectral feature is extracted to obtain current unbalance degree and current harmonic distortion rate, which constitutes a current characteristic parameter set; the stator winding temperature signal is filtered and denoised and trend is removed to obtain a processed signal; based on the processed signal, a temperature abnormal point is identified by a three-dimensional Euclidean distance k-neighbor clustering algorithm, and a temperature instantaneous change rate is calculated according to the temperature abnormal point to obtain a temperature characteristic parameter set containing abnormal point density and temperature rise rate; The current characteristic parameter set and the temperature characteristic parameter set are combined to construct a multi-dimensional dynamic feature vector; a dynamic evaluation domain is constructed according to the dynamic feature index; the dynamic evaluation domain is partitioned by grid to obtain a plurality of evaluation units; based on the distribution dispersion and time sequence evolution characteristics of the feature values in the evaluation units, a working condition self-adaptive adjustment coefficient is calculated; the current and temperature characteristic parameter sets are fused, and the initial evaluation value of the stator insulation state is calculated by a weighted comprehensive evaluation method; the initial evaluation value is dynamically corrected by the working condition self-adaptive adjustment coefficient to obtain the final stator insulation state evaluation result; The final stator insulation state evaluation result is compared with the preset safety threshold interval, and a differentiated stator fault early warning signal is generated according to the deviation level.

[0006] Further, through the current sensor and the temperature sensor, the three-phase current signal and the stator winding temperature signal during the operation of the motor are synchronously collected in real time, including: The data acquisition unit at the motor terminal issues a synchronous collection instruction, and the data acquisition unit controls the current sensor and the temperature sensor according to the collection instruction to trigger synchronous sampling of the three-phase current signal and the stator winding temperature signal, thereby obtaining synchronous sampled data; The data stream obtained by synchronous sampling is written into a terminal pre-allocated ring memory buffer for temporary caching, and the multi-channel data in the cache is packed into data frames and synchronized with time stamps according to the sampling time sequence; The data frames that have completed time stamp alignment are subjected to data quality verification and signal preprocessing to obtain preprocessed standardized data, thereby completing real-time synchronous collection of the three-phase current signal and the stator winding temperature signal of the motor.

[0007] Further, the three-phase current signal is analyzed in time-frequency domain and spectral feature is extracted to obtain current unbalance degree and current harmonic distortion rate, which constitutes a current characteristic parameter set, including: The three-phase current data is subjected to digital filtering and denoising to eliminate high-frequency sampling noise and conduction interference; The preprocessed current data is subjected to joint analysis in time-frequency domain, which includes calculating the amplitude and phase information of the fundamental wave and each harmonic component of the current signal by fast Fourier transform to obtain spectral analysis results; Based on the spectral analysis results, the unbalance degree index and the total harmonic distortion rate index of the three-phase current are calculated; The calculated current imbalance and current harmonic distortion rate are taken as core characteristic parameters to form a characteristic parameter set for representing the electrical state of the motor.

[0008] Further, the stator winding temperature signal is filtered and denoised and trend removed to obtain a processed signal; based on the processed signal, a temperature anomaly point is identified by a k-nearest neighbor clustering algorithm of three-dimensional Euclidean distance, and a temperature instantaneous change rate is calculated according to the temperature anomaly point to obtain a temperature characteristic parameter set containing anomaly point density and temperature rise rate, including: The collected stator winding temperature signal is digitally filtered to suppress high-frequency noise and power frequency interference to obtain a preliminary filtered temperature signal; the preliminary filtered temperature signal is subjected to trend removal processing, and the long-term trend component in the signal is removed by a moving average method to obtain a de-trended temperature signal; Based on the de-trended temperature signal, a three-dimensional feature space including time points, temperature values and temperature change rates is constructed; in the three-dimensional feature space, the Euclidean distance of each data point is calculated by a k-nearest neighbor clustering algorithm, and the temperature anomaly point is identified according to the distance threshold; According to the identified temperature anomaly point, a temperature instantaneous change rate is calculated, wherein the temperature instantaneous change rate is obtained by temperature difference operation in a time window near the anomaly point; The frequency of occurrence of temperature anomaly points per unit time is counted to obtain an anomaly point density index; at the same time, a temperature rise rate index is calculated based on the temperature instantaneous change rate; The anomaly point density and temperature rise rate are taken as core characteristic parameters to form a characteristic parameter for representing the stator thermal state.

[0009] Further, the current characteristic parameter set and the temperature characteristic parameter set are combined to construct a multi-dimensional dynamic feature vector; a dynamic evaluation domain is constructed according to the dynamic characteristic index; the dynamic evaluation domain is subjected to grid partitioning to obtain a plurality of evaluation units; based on the distribution dispersion and time sequence evolution characteristics of the feature values in the evaluation units, a working condition self-adaptive adjustment coefficient is calculated, including: The current characteristic parameter set and the temperature characteristic parameter set are data fused and organized according to a preset time sequence window to construct a feature vector containing multi-dimensional indexes of electrical and thermal states; Based on the numerical range of the multi-dimensional dynamic feature vector, a dynamic change interval is defined for each feature dimension to construct a multi-dimensional dynamic evaluation domain; The multi-dimensional dynamic evaluation domain is subjected to uniform grid partitioning processing, and the entire evaluation domain is divided into a plurality of evaluation units with equal volume; For each evaluation unit, the number of historical and real-time feature vectors falling into the evaluation unit is counted, and the distribution dispersion of these feature values in the unit is calculated to obtain a distribution dispersion calculation result; analyzing the time sequence evolution characteristics of the feature vectors in each evaluation unit, and representing the time sequence evolution characteristics by calculating the moving path and the aggregation state change of the feature vectors in adjacent time sequence windows to obtain time sequence evolution characteristic analysis results; The distribution dispersion calculation results and the time sequence evolution characteristic analysis results of each evaluation unit are integrated, and a global operating condition adaptive adjustment coefficient is calculated by a weighted fusion method.

[0010] Further, the current and temperature feature parameter sets are fused, and the initial evaluation value of the stator insulation state is calculated by a weighted comprehensive evaluation method; the initial evaluation value is dynamically corrected by the operating condition adaptive adjustment coefficient to obtain the final stator insulation state evaluation result, including: Based on the current feature parameter set and the temperature feature parameter set, the reference values of the feature parameters are extracted from the current feature parameter set and the temperature feature parameter set respectively, and the initial weight coefficients of each feature parameter are determined by the analytic hierarchy process according to the degree of influence of the parameters on the insulation state; Based on the determined initial weight coefficients, each feature parameter is weighted and processed to calculate the weighted evaluation value of each feature parameter; the weighted evaluation values of the feature parameters are linearly superimposed to obtain a preliminary comprehensive evaluation value; The preliminary comprehensive evaluation value is normalized to map the preliminary comprehensive evaluation value to a preset evaluation interval to obtain the initial evaluation value of the stator insulation state; Based on the initial evaluation value of the stator insulation state, the initial evaluation value is dynamically corrected by the operating condition adaptive adjustment coefficient to obtain the dynamic correction processing result of the stator insulation state evaluation.

[0011] Further, the final stator insulation state evaluation result is compared with the preset safety threshold interval, and a differential stator fault warning signal is generated according to the deviation level, including: The stator insulation state evaluation result is taken as the insulation state quantitative value of the current evaluation period; based on the obtained insulation state quantitative value, the preset multi-level safety threshold interval configuration parameters are retrieved; The current insulation state quantitative value is compared with the retrieved safety threshold intervals level by level to determine the specific safety level interval in which the quantitative value is located; According to the determined safety level interval, the corresponding deviation level is mapped; based on the mapped deviation level, the differential stator fault warning signal of the corresponding level is obtained.

[0012] In a second aspect, a motor stator fault detection system includes: The acquisition module is configured to acquire three-phase current signals and stator winding temperature signals in real time and synchronously during motor operation through current sensors and temperature sensors. The computing module is used for time-frequency domain analysis and spectrum feature extraction on the three-phase current signal, so as to obtain a current unbalance degree and a current harmonic distortion rate, and form a current characteristic parameter set; the stator winding temperature signal is subjected to filtering and noise reduction and trend elimination processing, so as to obtain a processed signal; based on the processed signal, a temperature abnormal point is identified through a three-dimensional Euclidean distance k-neighbor clustering algorithm, a temperature instantaneous change rate is calculated according to the temperature abnormal point, and a temperature characteristic parameter set containing an abnormal point density and a temperature rise rate is obtained; The evaluation module is used for combining the current characteristic parameter set and the temperature characteristic parameter set, constructing a multi-dimensional dynamic characteristic vector, constructing a dynamic evaluation domain according to the dynamic characteristic index, carrying out grid partitioning on the dynamic evaluation domain, obtaining a plurality of evaluation units, calculating an operating condition self-adaptive adjustment coefficient based on the distribution discrete degree and the time sequence evolution characteristic of the characteristic value in the evaluation unit, fusing the current and temperature characteristic parameter sets, calculating an initial evaluation value of the stator insulation state through a weighted comprehensive evaluation mode, and dynamically correcting the initial evaluation value through the operating condition self-adaptive adjustment coefficient to obtain a final stator insulation state evaluation result. The processing module is used for comparing the final stator insulation state evaluation result with a preset safety threshold interval, and generating a differential stator fault early warning signal according to a deviation level.

[0013] In a third aspect, a computing device includes: one or more processors; a memory device storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0014] In a fourth aspect, a computer readable storage medium stores a program, when the program is executed by a processor, the method is implemented.

[0015] The above scheme of the present application at least has the following beneficial effects: The technical means is that, by adopting the current sensor and the temperature sensor to synchronously collect the three-phase current and the stator winding temperature signal in real time, the current signal is analyzed in time domain and frequency domain to extract the current unbalance degree and the harmonic distortion rate, the temperature signal is filtered and de-trended, then the abnormal points are identified by three-dimensional Euclidean distance k neighbor clustering and the abnormal point density and the temperature rise rate are calculated, then the two types of features are combined to construct a multi-dimensional dynamic feature vector, the dynamic evaluation domain is gridded and partitioned, the self-adaptive adjustment coefficient of the working condition is calculated based on the feature distribution dispersion and the time sequence evolution characteristics in the unit, the feature weight is determined by the analytic hierarchy process, the initial evaluation value is obtained by weighted comprehensive evaluation and normalization, finally the evaluation value is dynamically corrected combined with the self-adaptive adjustment coefficient of the working condition, and the differential early warning is generated by comparing with the multi-level safety threshold, so that the technical problems that the traditional detection method only relies on single parameter monitoring, lacks multi-parameter collaborative analysis capability, and is difficult to distinguish normal working condition fluctuation and fault signal, has weak anti-environmental interference ability and insufficient sensitivity to local faults, does not consider the fixed threshold caused by working condition change, and has high false alarm rate and high risk of missing report are overcome, and then the accurate identification of the motor stator fault is realized, the false alarm rate and the missing alarm rate are reduced, the working condition changes such as real-time adaptation to load fluctuation can be realized, different levels of faults can be disposed in time through differential early warning, the stable operation of the motor is ensured, the equipment burning and production interruption caused by faults are avoided, and the operation reliability and production continuity of the industrial system are improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of a motor stator fault detection method provided by an embodiment of the present application.

[0017] Figure 2 is a schematic diagram of a motor stator fault detection system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0019] As Figure 1 shown, an embodiment of the present application proposes a motor stator fault detection method, which comprises the following steps: Step 1, by the current sensor and the temperature sensor, synchronously collecting the three-phase current signal and the stator winding temperature signal when the motor is running in real time; Step 2, time-frequency domain analysis and spectrum feature extraction are performed on the three-phase current signal to obtain current unbalance degree and current harmonic distortion rate, and a current characteristic parameter set is formed; the stator winding temperature signal is filtered and denoised and trend is removed to obtain a processed signal; based on the processed signal, a temperature abnormal point is identified through a three-dimensional Euclidean distance k-neighbor clustering algorithm, and a temperature instantaneous change rate is calculated according to the temperature abnormal point, and a temperature characteristic parameter set containing abnormal point density and temperature rise rate is obtained; Step 3, the current characteristic parameter set and the temperature characteristic parameter set are combined to construct a multi-dimensional dynamic feature vector; a dynamic evaluation domain is constructed according to the dynamic feature index; the dynamic evaluation domain is grid partitioned to obtain a plurality of evaluation units; a working condition adaptive adjustment coefficient is calculated based on the distribution dispersion degree and time sequence evolution characteristics of the feature values in the evaluation units; the initial evaluation value of the stator insulation state is calculated by weighted comprehensive evaluation method by fusing the current and temperature characteristic parameter sets; the initial evaluation value is dynamically corrected by the working condition adaptive adjustment coefficient to obtain the final stator insulation state evaluation result; Step 4, the final stator insulation state evaluation result is compared with the preset safety threshold interval, and a differential stator fault early warning signal is generated according to the deviation level.

[0020] In the embodiment of the present application, the three-phase current and the stator winding temperature signal are synchronously collected by the current sensor and the temperature sensor in real time, the current signal is analyzed in time-frequency domain to extract the current unbalance degree and the current harmonic distortion rate to form a current characteristic parameter set, the temperature signal is filtered and denoised and trend is removed, and then the temperature abnormal point is identified by the three-dimensional Euclidean distance k-neighbor clustering algorithm and the abnormal point density and the temperature rise rate are calculated to form a temperature characteristic parameter set, and then the two types of parameter sets are combined to construct a multi-dimensional dynamic feature vector, the dynamic evaluation domain is grid partitioned, and the working condition adaptive adjustment coefficient is calculated based on the feature distribution dispersion degree and the time sequence evolution characteristics in the evaluation units, and the initial evaluation value is obtained by weighted comprehensive evaluation, and then the initial evaluation value is dynamically corrected by the working condition adaptive adjustment coefficient, and finally the result is compared with the preset safety threshold interval to generate a differential early warning technical means, so that the technical problems of the traditional detection method, such as relying on single parameter monitoring, lacking of multi-parameter collaborative analysis ability, being difficult to distinguish normal working condition fluctuation and fault signal, weak anti-environmental interference ability, insufficient sensitivity to local fault, fixed threshold value caused by considering working condition change, high false alarm rate and high risk of missing report, are overcome, and the motor stator fault is identified, the false alarm rate and the missing report rate are reduced, the working condition changes such as motor start-stop and load switching are adapted in real time, the differential early warning is used to timely deal with different levels of faults, the stable operation of the motor is effectively ensured, the equipment burning and production interruption caused by stator fault are avoided, and the operation reliability and continuity of the industrial production system are improved.

[0021] In a preferred embodiment of the present application, the above step 1 can comprise: Step 1.1, the motor terminal based on the data acquisition unit issued synchronous acquisition instruction, data acquisition unit according to the acquisition instruction control current sensor and temperature sensor, trigger three phase current signal and stator winding temperature signal synchronous sampling, get synchronous sampling data, including: the motor terminal is equipped with a special data acquisition unit, the data acquisition unit first according to the current running state of the motor and detection requirements, determine the three phase current signal and stator winding temperature signal synchronous acquisition time node and acquisition frequency; then, the data acquisition unit to itself issued synchronous acquisition instruction, the instruction contains the unified time requirements of starting collection; data acquisition unit receives the synchronous acquisition instruction, immediately send acquisition start signal to the connected current sensor and temperature sensor, respectively, to ensure that the current sensor and temperature sensor start working at the same time; wherein, the current sensor for the three phase winding of the motor respectively collect real-time current signal, temperature sensor then paste stator winding surface or into the winding internal collection real-time temperature signal, finally get the same time point corresponding to three phase current data and stator winding temperature data.

[0022] Step 1.2, the data stream obtained by synchronous sampling is written into the terminal pre allocated ring memory buffer for temporary caching, and the multi channel data in the cache is packed and time stamped synchronous alignment according to the sampling time sequence, including: after the completion of the synchronous sampling operation, the collected three phase current data stream and stator winding temperature data stream are transmitted to the motor terminal one by one according to the order of collection. The motor terminal pre allocates a special ring memory buffer for real time collection data. The buffer has a circular storage function, which can avoid data overflow or loss during temporary storage. After the data stream is written into the ring memory buffer for temporary storage, the three phase current data and temperature data belonging to the same collection time are extracted from the ring memory buffer according to the specific time sequence of each data point collection. The data at the same time are integrated together to form a complete data frame, and a unique and accurate time mark is added to each data frame. By comparing and adjusting the time mark of each data frame, it is ensured that each group of three phase current data and the corresponding temperature data are completely matched in time dimension, realizing the time stamp synchronous alignment of multi channel data.

[0023] Step 1.3, data quality check and signal preprocessing are performed on the data frame aligned with the completion timestamp, to obtain preprocessed standardized data, and real-time synchronous acquisition of motor three-phase current signals and stator winding temperature signals is completed, specifically including: for all data frames that have completed timestamp synchronization alignment, first, data quality check is carried out, and the structure of each data frame is checked one by one to confirm whether there are problems such as data loss and data misplacement in the data frame; at the same time, referring to the reasonable range of three-phase current and stator winding temperature when the motor is running normally, it is judged whether the current value and temperature value in each data frame are within the normal physical interval, and the data frame with structural defects or abnormal value is determined as invalid data and is eliminated; for the valid data frame passing the quality check, further signal preprocessing operation is performed to remove high-frequency noise signals in the data caused by external electromagnetic interference, sensor slight vibration and other factors, correct the slight value deviation in the data acquisition process, and convert the current data and temperature data of different formats into a standard data format recognizable by the motor terminal, so that the value range and representation form of all data remain consistent; after the above operation, the preprocessed standardized data is obtained, and thus the real-time synchronous acquisition of motor three-phase current signals and stator winding temperature signals is completed.

[0024] In the embodiment of the application, the synchronous acquisition instruction is issued based on the motor terminal data acquisition unit to control the current sensor and the temperature sensor to realize synchronous sampling of the three-phase current signal and the stator winding temperature signal, the synchronous sampling data stream is written into the pre-allocated ring memory buffer for temporary caching and multi-channel data frame packaging and timestamp synchronization alignment are completed according to the sampling time sequence, and data quality check and signal preprocessing are performed on the timestamp-aligned data frame, so that the problems of asynchronous sampling of sensors, time sequence misplacement of current and temperature data, easy loss of data caching, mixing of abnormal data due to lack of quality check in the traditional data acquisition process are overcome, the time sequence consistency of the three-phase current and the stator winding temperature data is ensured, data loss and abnormal data interference are avoided, and finally the preprocessed standardized synchronous data is obtained.

[0025] In a preferred embodiment of the application, the above step 2 can include: Step 2.1, digital filtering and denoising processing are performed on the three-phase current data to eliminate high-frequency sampling noise and conduction interference, specifically including: obtaining the three-phase current raw data obtained after synchronous acquisition, selecting an appropriate digital filtering method in combination with the current signal frequency range when the motor is running normally, setting filtering parameters to filter out high-frequency sampling noise higher than the rated running frequency of the motor, and eliminating conduction interference by suppressing the noise signal mixed in the power grid transmission process, real-time monitoring the change trend of the current data during the filtering process to ensure that the filtering operation does not lose the effective feature information in the current signal, and finally obtaining clean three-phase current data after removing noise and interference.

[0026] Step 2.2, performing time-frequency domain joint analysis on the preprocessed current data, the analysis including calculating the amplitude and phase information of the fundamental wave and each harmonic component of the current signal by fast Fourier transform to obtain the frequency spectrum analysis result, specifically including: the filtered and denoised three-phase current data after preprocessing, the filtered three-phase current data segment is preprocessed according to the preset time window, according to the operation requirement of fast Fourier transform, the data segment length is supplemented to the integer power of 2, such as the original data segment length is 1000 points, the zero is supplemented to 1024 points, to avoid the decrease of operation precision caused by the mismatch of data length, and at the same time to ensure that the time span of each data can completely cover at least one running cycle of the motor; then, performing fast Fourier transform operation on the preprocessed current data segment, the fast Fourier transform operation will convert the current signal changing with time in time domain into the data in frequency domain in the form of complex number, the real part and the imaginary part of the complex number correspond to the amplitude component and the phase component of the signal at different frequency points, so as to complete the conversion from observing the current change in time dimension to decomposing the current in frequency dimension; determining the frequency position of the fundamental wave and each harmonic, first determining the fundamental wave frequency according to the rated operating frequency of the motor, and then determining the 3rd, 5th and 7th harmonic frequencies according to 3 times, 5 times and 7 times the fundamental wave frequency; combining the sampling frequency of the current data and the operation point number of the fast Fourier transform, the actual frequency value corresponding to each frequency point in the frequency domain data is calculated, the frequency points completely matched or closest to the fundamental wave and each harmonic frequency are found, and the frequency domain complex data corresponding to these frequency points are taken as the target analysis object, then the amplitude is calculated, for the frequency domain complex data corresponding to the found fundamental wave and each harmonic, the preliminary amplitude is obtained by the modulus calculation method of complex number, and then the amplitude scaling rule of fast Fourier transform is used, such as single channel signal needs to be divided by data segment length, or needs to be multiplied by 2 for correction when only taking the positive frequency part due to the symmetry of frequency domain data, finally the fundamental wave and each harmonic amplitude consistent with the actual amplitude of the current signal is obtained; finally, the phase is calculated, based on the frequency domain complex data corresponding to the fundamental wave and each harmonic, the phase angle is determined by the amplitude angle calculation method of complex number, taking the real axis of complex plane as the reference, the value relationship between the imaginary part and the real part is determined, and the range of the phase angle is usually adjusted to- to , so as to reflect the signal phase state of the fundamental wave and each harmonic at the frequency point, such as the phase difference can reflect the time delay difference of different harmonic components; arranging all the calculated amplitude and phase data of the fundamental wave and each harmonic in order from low to high frequency to form a complete current signal frequency spectrum analysis result.

[0027] Step 2.3, based on the spectrum analysis result, the unbalance degree index of three-phase current and the total harmonic distortion rate index are calculated, specifically including: extracting the fundamental wave amplitude data of each three-phase current from the spectrum analysis result, first calculating the average value of the three-phase current fundamental wave amplitude, then calculating the deviation of each phase current fundamental wave amplitude from the average value, finding the maximum deviation value, and dividing the maximum deviation value by the average value to obtain the unbalance degree index of the three-phase current; at the same time, according to the amplitude data of each harmonic in the spectrum analysis result, the square sum of all harmonic component amplitudes is calculated and square root is taken to obtain the total harmonic amplitude, and the total harmonic amplitude is divided by the fundamental wave amplitude to obtain the total harmonic distortion rate index, and the specific values of the two indexes are recorded.

[0028] Step 2.4, the calculated current unbalance degree and current harmonic distortion rate are used as core characteristic parameters to form a characteristic parameter set for representing the electrical state of the motor, specifically including: classifying and arranging the calculated three-phase current unbalance degree index and total harmonic distortion rate index, and storing the corresponding values of the two indexes in each sampling period in chronological order of the motor operation to form a group of core parameters that can fully reflect the electrical operation state of the motor; the two parameters are used to judge whether the motor has the problems of increased harmonic components caused by stator winding inter-turn short circuit and current imbalance caused by asymmetric three-phase winding impedance, and finally form a characteristic parameter set for accurately representing the electrical state of the motor.

[0029] In the embodiment of the application, the three-phase current data is subjected to digital filtering and denoising to eliminate high-frequency sampling noise and conducted interference, time-frequency domain joint analysis is performed on the preprocessed current data, and the amplitude and phase information of the fundamental wave and each harmonic component of the current signal are calculated by fast Fourier transform, based on the spectrum analysis result, the unbalance degree index of three-phase current and the total harmonic distortion rate index are calculated, and the two indexes are used as core characteristic parameters to form a characteristic parameter set for representing the electrical state of the motor, so that the technical problems of traditional detection that three-phase current data is easily distorted by high-frequency noise and conducted interference, and only relying on a single current parameter cannot capture the electrical abnormal characteristics of the motor and cannot distinguish between normal load fluctuations and real current fault signals are overcome, and clean and reliable current analysis data is obtained, the core characteristic parameters that can fully reflect the electrical operation state of the motor are extracted, and the risk of fault misjudgment caused by data distortion or single parameter is reduced from the electrical characteristic extraction level.

[0030] In a preferred embodiment of the application, step 2 can include: Step 2.5, the collected stator winding temperature signal is digitally filtered to suppress high-frequency noise and power frequency interference, and a preliminary filtered temperature signal is obtained; the preliminary filtered temperature signal is subjected to trend elimination processing, and the long-term trend component in the signal is removed through a moving average method to obtain a de-trended temperature signal, which specifically includes: obtaining the synchronously collected stator winding temperature original signal, combining the characteristics of high-frequency noise and power frequency interference in the temperature signal, selecting an appropriate digital filtering method, setting filtering parameters to specifically suppress these interference signals, ensuring that the filtered temperature signal can retain the true temperature change trend, and obtaining a preliminary filtered temperature signal; then, according to the characteristics of the slow change of the environmental temperature in the motor operating environment and the gradual drift of the temperature when the motor load is long-term stable, the window length of the moving average method is determined, the window length needs to cover the change period of the long-term trend to avoid excessive smoothing and loss of effective temperature fluctuations, the average value of the preliminary filtered temperature signal is calculated in segments according to the window length, and the average value of the corresponding window is subtracted from the original temperature signal to remove the long-term trend component in the signal, and finally a de-trended temperature signal reflecting the instantaneous change of the temperature of the stator winding itself is obtained.

[0031] Step 2.6, based on the de-trended temperature signal, a three-dimensional feature space including time points, temperature values and temperature change rates is constructed; in the three-dimensional feature space, the Euclidean distance of each data point is calculated using a k-nearest neighbor clustering algorithm, and temperature abnormal points are identified according to a distance threshold, which specifically includes: based on the de-trended temperature signal, the time point corresponding to each sampling time, the temperature value of each sampling time, and the temperature difference between adjacent two sampling times are extracted, and then divided by the time interval of the two times to obtain the temperature change rate corresponding to each sampling time; the time point, temperature value and temperature change rate of each sampling time are taken as a group to construct a three-dimensional feature space, and each data point in the space can reflect the time attribute, numerical attribute and change attribute of the temperature of the stator winding at a certain time; in the three-dimensional feature space, first, based on the historical temperature data of the normal operation of the motor, the k value of the k-nearest neighbor clustering algorithm is determined, the k value needs to be calibrated through multiple tests to ensure that normal data and abnormal data can be accurately distinguished, then the Euclidean distance of each data point and the k adjacent data points is calculated, and the range of the Euclidean distance of the data points under normal operating conditions is counted to set the distance threshold; if the Euclidean distance of a data point and the k adjacent data points exceeds the threshold, the data point is determined to be a temperature abnormal point, and the identification of the temperature abnormal point is completed.

[0032] Step 2.7. According to the identified temperature abnormal points, the temperature instantaneous change rate is calculated, wherein the temperature instantaneous change rate is obtained by difference operation of the temperature in the time window near the abnormal point, and specifically includes: for each identified temperature abnormal point, first determine the size of the time window near the abnormal point according to the sampling frequency of the temperature signal, the time window needs to contain several sampling time points before and after the abnormal point, to ensure that the complete temperature change process around the abnormal point is covered, and to avoid insufficient data and large calculation error caused by too small window; extract the temperature values of all sampling time points in the time window from the detrended temperature signal, calculate the temperature difference of adjacent sampling time points in the window, and then divide by the time interval of adjacent time points to obtain multiple temperature change rate data in the window; filter these temperature change rate data to remove extreme values that deviate significantly from the normal range, and take the average or median of the remaining data as the temperature instantaneous change rate corresponding to the temperature abnormal point, to ensure that the calculation result can truly reflect the instantaneous change of the temperature at the abnormal point.

[0033] Step 2.8. The frequency of temperature abnormal points in a unit of time is counted to obtain the abnormal point density index; at the same time, the temperature rise rate index is calculated based on the temperature instantaneous change rate, which specifically includes: first set the time unit for counting the abnormal point density, such as 1 minute or 5 minutes, the time unit needs to be set in combination with the motor fault development speed, to capture the abnormal trend in time and avoid large fluctuations in the statistical result caused by too small unit; in the time unit, the total number of identified temperature abnormal points is counted, and the total number is divided by the length of the time unit to obtain the frequency of temperature abnormal points in a unit of time, i.e. the abnormal point density index, which can reflect the frequency of temperature abnormalities of the stator winding; at the same time, the temperature instantaneous change rate data corresponding to all temperature abnormal points are collected, and the average or maximum value of the remaining data is calculated after removing the extreme abnormal values, as the temperature rise rate index, which can reflect the speed of temperature abnormal rise of the stator winding.

[0034] Step 2.9. The abnormal point density and the temperature rise rate are taken as core feature parameters to form a feature parameter set for representing the thermal state of the stator, which specifically includes: the calculated abnormal point density index and temperature rise rate index are stored in chronological order, each time unit corresponds to a set of abnormal point density and temperature rise rate data; in combination with the historical data of the normal operation of the motor, the numerical range of the abnormal point density index and the temperature rise rate index in the normal state is determined, and the current thermal state of the stator winding can be quickly judged by comparison; the abnormal point density index and the temperature rise rate index are taken as core feature parameters to form a complete feature parameter set, which can comprehensively and accurately represent the thermal state of the stator winding from two dimensions of the frequency of temperature abnormalities and the change speed of temperature abnormalities.

[0035] In the embodiment of the present application, the collected stator winding temperature signal is first subjected to digital filtering to suppress high-frequency noise and power frequency interference, and then the trend elimination processing is performed by the moving average method to remove the long-term trend component, the three-dimensional feature space containing the time point, temperature value and temperature change rate is constructed based on the detrended temperature signal, and the k-nearest neighbor clustering algorithm is used to calculate the Euclidean distance of each data point to identify the temperature abnormal point, the temperature instantaneous change rate is calculated through the temperature difference operation in the nearby time window according to the identified abnormal point, the abnormal point density index is obtained by counting the frequency of abnormal points in unit time, and the temperature rise rate index is calculated based on the temperature instantaneous change rate, and finally the two indexes are used as the core feature parameters to form the feature parameter set representing the stator thermal state, so that the technical problems of the traditional stator temperature monitoring, such as the distortion of temperature signal caused by high-frequency noise, power frequency interference and air flow in workshop, the difficulty in removing long-term trend interference and accurately identifying local temperature abnormal points, and the lack of feature parameters reflecting the dynamic change of thermal state, which leads to the insufficient sensitivity to stator winding local micro-short circuit fault and the risk of false negatives, are overcome, and the clean temperature signal without trend interference is obtained, the accurate capture of local temperature abnormal points of stator winding is realized, and the core feature parameters reflecting the change trend of stator thermal state are extracted, thereby effectively reducing the risk of fault false negatives caused by inaccurate temperature monitoring.

[0036] In a preferred embodiment of the present application, step 3 can include: Step 3.1, data fusion is performed on the current feature parameter set and the temperature feature parameter set, and a feature vector containing multi-dimensional indexes of electrical and thermal states is constructed according to a preset time sequence window, which specifically includes: extracting current unbalance degree and current harmonic distortion rate from the obtained current feature parameter set, and extracting abnormal point density and temperature rise rate from the temperature feature parameter set; according to the working condition characteristics of the motor in the new energy automobile power battery production line, the preset time sequence window is set to 15 seconds to ensure that each window can cover a short-term load fluctuation period; in each 15-second time sequence window, the average value of current unbalance degree, the average value of current harmonic distortion rate, the average value of abnormal point density and the average value of temperature rise rate are calculated; in the fixed order of current unbalance degree, current harmonic distortion rate, abnormal point density and temperature rise rate, the four average values are integrated into a group of data to form a multi-dimensional dynamic feature vector containing information of electrical operation state of the motor and thermal state of the stator, and each vector corresponds to the comprehensive operation of the motor in a time sequence window.

[0037] Step 3.2, based on the numerical range of the multi-dimensional dynamic feature vector, define the dynamic change interval for each feature dimension, construct a multi-dimensional dynamic evaluation domain, specifically including: collect all multi-dimensional dynamic feature vectors generated by the motor during normal operation in the past four months, respectively statistics of the actual numerical range of each feature dimension: the current unbalance degree is in the numerical range of 0.3% to 4% during normal operation, considering that load switching may cause short-term numerical increase, the dynamic change interval is extended to 0 to 12%; the current harmonic distortion rate is in the numerical range of 0.2% to 3% during normal operation, the dynamic change interval is extended to 0 to 10%; the abnormal point density is in the numerical range of 0 to 0.3 per second during normal operation, the dynamic change interval is extended to 0 to 3 per second; the temperature rise rate is in the numerical range of 0 to 0.6℃ / min during normal operation, the dynamic change interval is extended to 0 to 6℃ / min; based on the dynamic change interval of the four feature dimensions, a multi-dimensional dynamic evaluation domain is built which can cover the normal operation state and potential fault state of the motor.

[0038] Step 3.3, uniform grid partitioning processing is performed on the multi-dimensional dynamic evaluation domain, and the entire evaluation domain is divided into several evaluation units with equal volume, specifically including: uniform segmentation processing is performed on each feature dimension of the multi-dimensional dynamic evaluation domain, the 0 to 12% interval of the current unbalance degree is uniformly divided into 12 segments, and each segment corresponds to a numerical interval of 1%; the 0 to 10% interval of the current harmonic distortion rate is uniformly divided into 10 segments, and each segment corresponds to a numerical interval of 1%; the 0 to 3 per second interval of the abnormal point density is uniformly divided into 15 segments, and each segment corresponds to a numerical interval of 0.2 per second; the 0 to 6℃ / min interval of the temperature rise rate is uniformly divided into 12 segments, and each segment corresponds to a numerical interval of 0.5℃ / min; through the above segmentation method, the entire multi-dimensional dynamic evaluation domain is segmented into several evaluation units with equal volume, for example, 12x10x15x12=21600 evaluation units, each evaluation unit corresponds to a small range of numerical combination of each feature dimension, which ensures that the specific position of each multi-dimensional dynamic feature vector in the evaluation domain can be accurately located.

[0039] Step 3.4, for each evaluation unit, statistics the number of historical and real-time feature vectors falling into the evaluation unit, and calculates the distribution dispersion of the feature values in the unit to obtain the distribution dispersion calculation result, specifically including: for each evaluation unit, first statistics the total number of historical multi-dimensional dynamic feature vectors falling into the evaluation unit in the past three months when the motor is running normally, and then statistics the number of real-time multi-dimensional dynamic feature vectors falling into the evaluation unit in the current sampling period, i.e. in the last 15-second time window, combine the historical vector number and the real-time vector number to form the total set of feature vectors of the evaluation unit; extract the specific values of each feature dimension from the total set of feature vectors, for example, extract the current unbalance degree values of all vectors, calculate the dispersion degree of the current unbalance degree values, realize by statistics the difference degree between the values, in the same way, respectively calculate the dispersion degree of the current harmonic distortion rate, the abnormal point density, the temperature rise rate; take the average value of the dispersion degree of the four feature dimensions as the distribution dispersion calculation result of the feature values in the evaluation unit, if the average value is small, it means that the feature value distribution in the unit is concentrated, which is consistent with the normal working condition characteristics; if the average value is large, it means that the feature value distribution is dispersed, there is working condition fluctuation or potential failure.

[0040] Step 3.5, analyze the time sequence evolution characteristics of the feature vectors in each evaluation unit, and represent by calculating the moving path and aggregation state change of the feature vectors in adjacent time sequence windows to obtain the time sequence evolution characteristics analysis result, specifically including: analyzing the time sequence evolution characteristics of the feature vectors in each evaluation unit, taking the adjacent two 15-second time sequence windows, i.e. the previous time sequence window and the current time sequence window as the analysis object, tracking the evaluation unit position of each multi-dimensional dynamic feature vector in the adjacent window: for example, a certain feature vector is located in evaluation unit X in the previous window and evaluation unit Y in the current window, calculate the spatial distance of evaluation unit X and evaluation unit Y in the multi-dimensional dynamic evaluation domain, the spatial distance is the moving path length of the feature vector; At the same time, statistics the feature vector number change ratio of each evaluation unit in the adjacent two time sequence windows, for example, there are 8 feature vectors in evaluation unit Z in the previous window, and there are 24 feature vectors in the current window, the number change ratio is 200%; judge the migration trend of the feature vector by the moving path length, judge the aggregation state change of the feature vector by the number change ratio, combine the two judgment results to form the time sequence evolution characteristics analysis result of each evaluation unit.

[0041] Step 3.6, the distribution dispersion calculation results and the time sequence evolution characteristic analysis results of each evaluation unit are integrated to obtain a global working condition adaptive adjustment coefficient through weighted fusion, and the specific steps include: in the comprehensive analysis, according to the operation experience of the motor in the power battery production line, the fusion weights of the distribution dispersion results and the time sequence evolution characteristic results are set as: the weight of the distribution dispersion results is 0.5, and the weight of the time sequence evolution characteristic results is 0.5; the distribution dispersion calculation results of each evaluation unit are multiplied by 0.5, the time sequence evolution characteristic analysis results are multiplied by 0.5, and the sum of the two is obtained. Local working condition adjustment coefficient of evaluation unit; then according to the total number of feature vectors in each evaluation unit, the sum of the historical vector quantity and the real-time vector quantity, the unit weight is set, the evaluation unit with more total number of feature vectors is more representative of the global working condition, and the unit weight is higher, and the specific weight value is the proportion of the total number of evaluation unit vectors in the total number of evaluation unit vectors; finally, the local working condition adjustment coefficient of each evaluation unit is multiplied by the corresponding unit weight, and the weighted results of all evaluation units are added and averaged to obtain the global working condition adaptive adjustment coefficient. The working condition adaptive adjustment coefficient can reflect the deviation degree of the current motor working condition from the normal working condition.

[0042] In the embodiment of the application, the current characteristic parameter set and the temperature characteristic parameter set are data fused according to a preset time sequence window to construct a feature vector containing multi-dimensional indexes of electrical and thermal states, a dynamic change interval is defined for each feature dimension based on the numerical range of the feature vector, a multi-dimensional dynamic evaluation domain is constructed, a plurality of evaluation units with equal volume are obtained by uniformly meshing the multi-dimensional dynamic evaluation domain, the number of historical and real-time feature vectors falling into each evaluation unit is counted, the distribution dispersion of feature values is calculated, the time sequence evolution characteristic of the feature vectors in each evaluation unit is analyzed, and finally the distribution dispersion calculation results and the time sequence evolution characteristic analysis results of each evaluation unit are integrated to obtain a global working condition adaptive adjustment coefficient through weighted fusion. Therefore, the technical problems of the traditional detection method, such as lack of multi-parameter collaborative fusion capability, no dynamic evaluation system established, and no consideration of the influence of working condition change on feature parameters, which lead to difficulty in distinguishing between motor normal load fluctuation and real fault signal and fixed detection threshold which cannot adapt to dynamic changes of working conditions, are overcome, and the comprehensive integration and dynamic tracking of multi-dimensional running information of electrical and thermal states of the motor are realized. The working condition adaptive adjustment coefficient can accurately reflect the influence of the current working condition on the feature parameters.

[0043] In a preferred embodiment of the application, step 3 can include: Step 3.7, based on the current characteristic parameter set and the temperature characteristic parameter set, the reference values of the characteristic parameters are extracted from the current characteristic parameter set and the temperature characteristic parameter set respectively, and the initial weight coefficients of each characteristic parameter are determined by the analytic hierarchy process according to the degree of influence of the parameters on the insulation state, which specifically includes: extracting the reference values of the current unbalance degree and the current harmonic distortion rate from the current characteristic parameter set, the reference values come from the historical data of the motor in the new energy automobile power battery production line within three months of normal operation, taking the upper limit of the 95% confidence interval as the reference, such as the current unbalance degree reference value is 3%, the current harmonic distortion rate reference value is 2%; the reference values of the abnormal point density and the temperature rise rate are extracted from the temperature characteristic parameter set, based on the same historical normal data, the abnormal point density reference value is 0.2 per second, and the temperature rise rate reference value is 0.5℃ / min; the initial weight coefficients are determined by the analytic hierarchy process, five motor operation and maintenance experts are invited, and the influence of each parameter on the insulation state of the stator is judged according to the characteristics of the power battery production line motor, such as the current harmonic distortion rate is more sensitive to turn-to-turn short circuit and has the greatest impact; the temperature rise rate affects the insulation aging, a judgment matrix is constructed and consistency test is carried out, and finally the current unbalance degree weight is 0.2, the current harmonic distortion rate weight is 0.3, the abnormal point density weight is 0.2, and the temperature rise rate weight is 0.3.

[0044] Step 3.8, based on the determined initial weight coefficients, each characteristic parameter is weighted and processed, and the weighted evaluation value of each characteristic parameter is calculated; the weighted evaluation value of the characteristic parameter is linearly superimposed to obtain a preliminary comprehensive evaluation value, which specifically includes: based on the determined initial weight coefficients, each characteristic parameter is weighted and processed: the deviation proportion of the current current unbalance degree and the reference value is calculated, such as the current value is 4%, the deviation proportion is (4%-3%) ÷ 3%, multiplied by the weight 0.2 to obtain the weighted evaluation value of the current unbalance degree; in the same way, the current harmonic distortion rate deviation proportion is multiplied by the weight 0.3, the abnormal point density deviation proportion is multiplied by the weight 0.2, and the temperature rise rate deviation proportion is multiplied by the weight 0.3, respectively to obtain the respective weighted evaluation value; the four weighted evaluation values are linearly superimposed, that is, the four values are added to obtain a preliminary comprehensive evaluation value.

[0045] Step 3.9, the initial evaluation value of the stator insulation state is obtained by mapping the preliminary comprehensive evaluation value into a preset evaluation interval through normalization processing of the preliminary comprehensive evaluation value, and specifically includes: collecting the preliminary comprehensive evaluation values when the motor stator fails in the past half year to determine the maximum value, such as 5; collecting the preliminary comprehensive evaluation values when the motor is running normally at the same period to determine the minimum value, such as 0; the preset evaluation interval is set to 0 to 1, and the preliminary comprehensive evaluation value is normalized by proportional conversion: subtract the minimum value from the current preliminary comprehensive evaluation value, and then divide by the difference between the maximum value and the minimum value, to obtain a value mapped into the interval of 0 to 1, which is the initial evaluation value of the stator insulation state. The closer the value is to 1, the worse the insulation state is.

[0046] Step 3.10, based on the initial evaluation value of the stator insulation state, the initial evaluation value is dynamically corrected by the working condition self-adaptive adjustment coefficient to obtain the stator insulation state evaluation result after dynamic correction, specifically including: obtaining the calculated global working condition self-adaptive adjustment coefficient, the coefficient range is 0.8 to 1.2, the coefficient greater than 1 indicates that the current working condition fluctuates greatly, which may amplify the parameter deviation; less than 1 indicates that the working condition is stable, and the parameter deviation is more reliable; based on the initial evaluation value of the stator insulation state, the working condition self-adaptive adjustment coefficient is dynamically corrected: when the adjustment coefficient is 1.1, the initial evaluation value is multiplied by 1.1 to appropriately enlarge the evaluation value to cope with the parameter abnormality caused by working condition fluctuation; when the adjustment coefficient is 0.9, the initial evaluation value is multiplied by 0.9 to appropriately reduce the evaluation value to avoid misjudgment of failure due to slight deviation when the working condition is stable; the value obtained after correction is the stator insulation state evaluation result after dynamic correction.

[0047] In the embodiment of the present application, the feature parameter reference values are extracted based on the current feature parameter set and the temperature feature parameter set, the initial weight coefficient of each feature parameter is determined by the analytic hierarchy process according to the difference of the influence of each parameter on the stator insulation state, the weighted processing is performed on each feature parameter based on the initial weight coefficient, and the weighted evaluation value is linearly superimposed to obtain the preliminary comprehensive evaluation value. The preliminary comprehensive evaluation value is normalized to map to a preset evaluation interval to obtain the initial evaluation value of the stator insulation state, and finally the initial evaluation value is dynamically corrected by the working condition self-adaptive adjustment coefficient. Therefore, the technical problems of the traditional detection method that does not distinguish the importance of different parameters on the insulation state, uses fixed weight or no weight calculation to cause the evaluation value to be unable to reflect the parameter priority, and does not correct the evaluation value combined with the working condition fluctuation to cause the evaluation value to deviate from the actual insulation state are overcome, and the stator insulation state evaluation result obtained can not only reflect the real influence of each feature parameter on the insulation state, but also adapt to the working condition changes such as motor start-stop and load switching, effectively reducing the fault misreporting or omission caused by distorted evaluation value.

[0048] In a preferred embodiment of the present application, step 4 above can include: Step 4.1, the stator insulation state evaluation result is taken as the insulation state quantitative value of the current evaluation period; based on the obtained insulation state quantitative value, the preset multi-level safety threshold interval configuration parameter is called, specifically including: the obtained dynamic corrected stator insulation state evaluation result is directly taken as the insulation state quantitative value of the current evaluation period, combined with the running rhythm of the new energy automobile power battery production line servo motor, the evaluation period is set to 1 minute, and the range of the insulation state quantitative value is 0 to 1; based on the obtained insulation state quantitative value, the multi-level safety threshold interval configuration parameter is called from the preset parameter library, wherein the parameter library is established combined with the historical fault data and operation and maintenance experience of the production line motor, and is specifically configured as a three-level safety threshold interval, the first level green normal interval is 0 to 0.4, the second level yellow potential risk interval is 0.4 to 0.7, and the third level red emergency fault interval is 0.7 to 1, so as to ensure that the motor covers the whole state range from normal operation to serious fault.

[0049] Step 4.2, the current insulation state quantitative value is compared with each safety threshold interval in turn to determine the specific safety level interval where the quantitative value is located, specifically including: in the order from high risk to low risk, the current insulation state quantitative value is compared with each safety threshold interval in turn: first, the red emergency fault interval is compared, if the quantitative value is greater than or equal to 0.7 and less than or equal to 1, it is determined that it is in the red emergency fault interval; if not, the yellow potential risk interval is continued to be compared, if the quantitative value is greater than or equal to 0.4 and less than 0.7, it is determined that it is in the yellow potential risk interval; if still not satisfied, the green normal interval is compared, if the quantitative value is greater than or equal to 0 and less than 0.4, it is determined that it is in the green normal interval; through this step-by-step investigation, the specific safety level interval where the current quantitative value is located is determined.

[0050] Step 4.3, according to the determined safety level interval, mapping to the corresponding deviation level; based on the deviation level obtained by mapping, the corresponding level difference is obtained. The stator fault warning signal, specifically including: according to the determined safety level interval, corresponding to the deviation level according to the preset mapping rule, when in the green normal interval, it is mapped to the no deviation level, indicating that the stator insulation state is good, and there is no fault risk; when in the yellow potential risk interval, it is mapped to the slight deviation level, indicating that the stator insulation state appears small abnormality, and there is potential fault risk; when in the red emergency fault interval, it is mapped to the serious deviation level, indicating that the stator insulation state is seriously abnormal, and there is clear fault sign; based on the deviation level obtained by mapping, the difference warning signal is generated: the no deviation level only generates the data record signal, and the current insulation state data is stored to the system database; the slight deviation level generates the sound and light alarm signal, triggers the warning light flickering and the buzzer alarm in the production line site, and pushes the warning prompt information to the terminal device of the operation and maintenance personnel; in addition to the sound and light alarm and terminal pushing signal, the serious deviation level also generates the load reduction control signal, reduces the servo motor load by 30% through the motor controller, avoids further expansion of the fault, and uploads the fault data to the production line central monitoring platform.

[0051] In the embodiment of the application, the stator insulation state evaluation result is taken as the insulation state quantitative value of the current evaluation period, the preset multi-level safety threshold interval configuration parameter is called based on the insulation state quantitative value, the current quantitative value is compared with each safety threshold interval to determine the specific safety level interval it is in, and then the safety level interval is mapped to the corresponding deviation level and the corresponding level difference stator fault warning signal is generated. Therefore, the technical problems of the traditional detection method, such as relying on a single fixed threshold to trigger the warning, being unable to distinguish the response level according to the insulation state risk degree, and being prone to false alarm under normal working conditions such as frequent start-stop or load switching of the motor, lack of targeted treatment for different severity faults, and the like are overcome, and the graded identification and differential response of the stator fault risk are realized, which can match the actual risk degree of the fault, avoid false alarm interference in production or missed treatment caused by a single threshold, improve the industrial scene adaptability of the fault warning, and ensure the safe operation of the motor and reduce the unplanned downtime loss.

[0052] As shown in Figure 2 The embodiment of the application also provides a motor stator fault detection system, which comprises: The acquisition module is used for synchronously collecting three-phase current signals and stator winding temperature signals in real time when the motor is running through the current sensor and the temperature sensor. The computing module is used for time-frequency domain analysis and spectrum feature extraction on the three-phase current signal, to obtain current unbalance degree and current harmonic distortion rate, and to form a current characteristic parameter set; the stator winding temperature signal is subjected to filtering and noise reduction and trend elimination processing, to obtain a processed signal; based on the processed signal, a temperature abnormal point is identified through a three-dimensional Euclidean distance k-neighbor clustering algorithm, a temperature instantaneous change rate is calculated according to the temperature abnormal point, and a temperature characteristic parameter set containing abnormal point density and temperature rise rate is obtained; The evaluation module is used for jointing the current characteristic parameter set and the temperature characteristic parameter set to construct a multi-dimensional dynamic characteristic vector; a dynamic evaluation domain is constructed according to the dynamic characteristic index; the dynamic evaluation domain is subjected to grid partitioning to obtain a plurality of evaluation units; based on the distribution dispersion and time sequence evolution characteristics of the characteristic values in the evaluation units, a working condition self-adaptive adjustment coefficient is calculated; the current and temperature characteristic parameter sets are fused, and an initial evaluation value of the stator insulation state is calculated through a weighted comprehensive evaluation method; the initial evaluation value is dynamically corrected through the working condition self-adaptive adjustment coefficient, to obtain a final stator insulation state evaluation result; The processing module is used for comparing the final stator insulation state evaluation result with a preset safety threshold interval, and generating a differential stator fault early warning signal according to the deviation level.

[0053] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles of the present application, can make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method of detecting a fault in a motor stator, the method comprising: The method comprises: Real-time synchronous acquisition of three-phase current signals and stator winding temperature signals during motor operation through current sensors and temperature sensors; Time-frequency domain analysis and spectral feature extraction are performed on the three-phase current signals to obtain current unbalance degree and current harmonic distortion rate, which constitute a current feature parameter set; the stator winding temperature signal is filtered and denoised and trend points are removed to obtain a processed signal; based on the processed signal, a three-dimensional Euclidean distance k-neighbor clustering algorithm is used to identify temperature abnormal points, and a temperature instantaneous change rate is calculated based on the temperature abnormal points to obtain a temperature feature parameter set containing abnormal point density and temperature rise rate; Jointly, the current feature parameter set and the temperature feature parameter set are used to construct a multi-dimensional dynamic feature vector; a dynamic evaluation domain is constructed based on the dynamic feature index; the dynamic evaluation domain is partitioned into a grid to obtain a plurality of evaluation units; based on the distribution dispersion and time sequence evolution characteristics of the feature values in the evaluation units, a working condition self-adaptive adjustment coefficient is calculated; the current and temperature feature parameter sets are fused, and the initial evaluation value of the stator insulation state is calculated by a weighted comprehensive evaluation method; the initial evaluation value is dynamically corrected by the working condition self-adaptive adjustment coefficient to obtain the final stator insulation state evaluation result; The final stator insulation state evaluation result is compared with the preset safety threshold interval, and a differentiated stator fault early warning signal is generated according to the deviation level.

2. A method of detecting a fault in a stator of an electrical machine according to claim 1, characterized in that, Real-time synchronous acquisition of three-phase current signals and stator winding temperature signals during motor operation through current sensors and temperature sensors, comprising: The data acquisition unit at the motor terminal issues a synchronous acquisition instruction, and the data acquisition unit controls the current sensor and the temperature sensor according to the acquisition instruction to trigger synchronous sampling of the three-phase current signal and the stator winding temperature signal, thereby obtaining synchronous sampled data; The data stream obtained by synchronous sampling is written into the terminal pre-allocated ring memory buffer for temporary caching, and the multi-channel data in the cache is packed into data frames and synchronized with time stamps according to the sampling time sequence; Data quality verification and signal preprocessing are performed on the data frames that have completed time stamp alignment to obtain preprocessed standardized data, thereby completing real-time synchronous acquisition of the motor three-phase current signal and the stator winding temperature signal.

3. A method of detecting a fault in a motor stator according to claim 2, characterized in that, Time-frequency domain analysis and spectral feature extraction are performed on the three-phase current signal to obtain current unbalance degree and current harmonic distortion rate, which constitute a current feature parameter set, comprising: Digital filtering and denoising are performed on the three-phase current data to eliminate high-frequency sampling noise and conducted interference; Time-frequency domain joint analysis is performed on the preprocessed current data, including calculating the amplitude and phase information of the fundamental wave and each harmonic component of the current signal using fast Fourier transform to obtain the spectral analysis result; Based on the spectral analysis result, the unbalance degree index and the total harmonic distortion rate index of the three-phase current are calculated; The calculated current unbalance degree and current harmonic distortion rate are used as core feature parameters to constitute a feature parameter set for representing the electrical state of the motor.

4. A method of detecting a fault in a motor stator according to claim 3, wherein Filtering and denoising and trend elimination are performed on the stator winding temperature signal to obtain a processed signal; based on the processed signal, a k-nearest neighbor clustering algorithm based on three-dimensional Euclidean distance is used to identify temperature abnormal points, and a temperature instantaneous change rate is calculated according to the temperature abnormal points to obtain a temperature characteristic parameter set containing abnormal point density and temperature rise rate, including: The collected stator winding temperature signal is subjected to digital filtering to suppress high-frequency noise and power frequency interference, and a preliminary filtered temperature signal is obtained; trend elimination is performed on the preliminary filtered temperature signal, and a moving average method is used to remove the long-term trend component in the signal to obtain a de-trended temperature signal; Based on the de-trended temperature signal, a three-dimensional feature space including time points, temperature values and temperature change rates is constructed; in the three-dimensional feature space, a k-nearest neighbor clustering algorithm is used to calculate the Euclidean distance of each data point, and temperature abnormal points are identified according to the distance threshold; According to the identified temperature abnormal points, the temperature instantaneous change rate is calculated, wherein the temperature instantaneous change rate is obtained by temperature difference operation in the time window near the abnormal point; The frequency of temperature abnormal points in a unit time is counted to obtain the abnormal point density index; at the same time, the temperature rise rate index is calculated based on the temperature instantaneous change rate; The abnormal point density and the temperature rise rate are taken as the core characteristic parameters to form the characteristic parameters for representing the stator thermal state.

5. A method of detecting a fault in a motor stator according to claim 4, wherein Combine the current characteristic parameter set and the temperature characteristic parameter set to construct a multi-dimensional dynamic feature vector; construct a dynamic evaluation domain according to the dynamic feature index; grid partition is performed on the dynamic evaluation domain to obtain a plurality of evaluation units; based on the distribution dispersion and time sequence evolution characteristics of the feature values in the evaluation unit, a working condition self-adaptive adjustment coefficient is calculated, including: Data fusion is performed on the current characteristic parameter set and the temperature characteristic parameter set, and a feature vector containing multi-dimensional indexes of electrical and thermal states is constructed according to a preset time sequence window; Based on the numerical range of the multi-dimensional dynamic feature vector, define the dynamic change interval for each feature dimension to construct a multi-dimensional dynamic evaluation domain; The multi-dimensional dynamic evaluation domain is uniformly grid partitioned to divide the entire evaluation domain into a plurality of evaluation units with equal volume; For each evaluation unit, the number of historical and real-time feature vectors falling into the evaluation unit is counted, and the distribution dispersion of these feature values in the unit is calculated to obtain the distribution dispersion calculation result; The time sequence evolution characteristics of the feature vector in each evaluation unit are analyzed by calculating the moving path and aggregation state change of the feature vector in adjacent time sequence windows to obtain the time sequence evolution characteristic analysis result; The distribution dispersion calculation result and the time sequence evolution characteristic analysis result of each evaluation unit are integrated, and a global working condition self-adaptive adjustment coefficient is calculated through weighted fusion.

6. A method of detecting a fault in a motor stator according to claim 5, wherein Fusion of current and temperature characteristic parameters is performed to calculate the initial evaluation value of the stator insulation state through a weighted comprehensive evaluation method; The initial evaluation value is dynamically corrected by the working condition self-adaptive adjustment coefficient to obtain the final stator insulation state evaluation result, including: Based on the current characteristic parameter set and the temperature characteristic parameter set, the reference values of the characteristic parameters are extracted from the current characteristic parameter set and the temperature characteristic parameter set respectively, and the initial weight coefficients of each characteristic parameter are determined by the analytic hierarchy process according to the degree of influence of the parameters on the insulation state; Based on the determined initial weight coefficients, each characteristic parameter is weighted and processed, and the weighted evaluation value of each characteristic parameter is calculated; the weighted evaluation values of the characteristic parameters are linearly superimposed to obtain a preliminary comprehensive evaluation value; The preliminary comprehensive evaluation value is normalized to map the preliminary comprehensive evaluation value into a preset evaluation interval to obtain an initial evaluation value of the stator insulation state; Based on the initial evaluation value of the stator insulation state, the initial evaluation value is dynamically corrected by the working condition adaptive adjustment coefficient to obtain a dynamically corrected stator insulation state evaluation result.

7. A method of detecting a fault in a motor stator according to claim 6, wherein The final stator insulation state evaluation result is compared with the preset safety threshold interval, and a differential stator fault early warning signal is generated according to the deviation level, including: The stator insulation state evaluation result is taken as the insulation state quantitative value of the current evaluation period; based on the obtained insulation state quantitative value, the preset multi-level safety threshold interval configuration parameter is called; The current insulation state quantitative value is compared with each safety threshold interval in turn to determine the specific safety level interval in which the quantitative value is located; According to the determined safety level interval, the corresponding deviation level is mapped; based on the mapped deviation level, a differential stator fault early warning signal of the corresponding level is obtained.

8. A motor stator fault detection system implementing the method of any one of claims 1 to 7, characterized in that, Including: The acquisition module is configured to acquire, through the current sensor and the temperature sensor, three-phase current signals and stator winding temperature signals in real time and synchronously when the motor is running; The calculation module is configured to perform time-frequency domain analysis and frequency spectrum feature extraction on the three-phase current signals to obtain current unbalance degree and current harmonic distortion rate, thereby forming a current characteristic parameter set; perform filtering and noise reduction and trend elimination processing on the stator winding temperature signals to obtain processed signals; and identify temperature abnormal points based on the processed signals by using a three-dimensional Euclidean distance k-nearest neighbor clustering algorithm, and calculate a temperature instantaneous change rate based on the temperature abnormal points, thereby obtaining a temperature characteristic parameter set containing abnormal point density and temperature rise rate; The evaluation module is configured to jointly use the current characteristic parameter set and the temperature characteristic parameter set to construct a multi-dimensional dynamic feature vector; construct a dynamic evaluation domain based on the dynamic feature index; perform grid partitioning on the dynamic evaluation domain to obtain a plurality of evaluation units; and calculate a working condition adaptive adjustment coefficient based on the distribution dispersion degree and time sequence evolution characteristics of the feature values in the evaluation units; The fusion of the current and temperature characteristic parameter sets is used to calculate an initial evaluation value of the stator insulation state by using a weighted comprehensive evaluation method; and the initial evaluation value is dynamically corrected by using the working condition adaptive adjustment coefficient to obtain a final stator insulation state evaluation result. The processing module is configured to compare the final stator insulation state evaluation result with a preset safety threshold interval, and generate a differential stator fault early warning signal according to a deviation level.

9. A computing device, comprising: Including: One or more processors; A storage device for storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program, which when executed by a processor, implements the method of any one of claims 1-7.

Images