Fault early warning method and system based on motor temperature monitoring

Abstract

The invention relates to the technical field of temperature monitoring, in particular to a fault early warning method and system based on motor temperature monitoring, and the method comprises the steps: obtaining operation data including the motor winding temperature, the environment temperature and a load current effective value; a temperature rise abnormal factor is calculated, and the temperature rise abnormal factor is in positive correlation with the difference value of the motor winding temperature and the environment temperature and is in negative correlation with the product of the square of the load current effective value and the preset motor thermal resistance. According to the method, a local density fluctuation factor is introduced to carry out adaptive correction on distance measurement in a traditional LOF (local anomaly factor) algorithm, and the local density is adjusted by using fluctuation stability (variance) in a data point neighborhood, so that data points which are not greatly deviated in a fault evolution period but are in an unstable state are endowed with lower density values; therefore, the sensitivity of the algorithm to the slow evolution type initial fault of the motor is obviously improved, and the missing report rate is reduced.

Description

Technical Field

[0001] This invention relates to the field of temperature monitoring technology. More specifically, this invention relates to a fault early warning method and system based on motor temperature monitoring. Background Technology

[0002] In industrial environments, motor temperature is a key indicator for monitoring motor operating status. However, temperature changes depend not only on the motor's internal condition, such as initial faults like bearing wear and winding insulation aging, but are also strongly influenced by external operating conditions and environmental factors. Existing fault warning systems, if based solely on raw temperature data for anomaly detection, are prone to misinterpreting normal load fluctuations or temperature increases caused by ambient temperature changes as faults, leading to a high false alarm rate. Furthermore, for slowly developing initial faults, the temperature change trend may be masked by environmental noise, resulting in a high false negative rate.

[0003] In data-driven fault diagnosis technology, the LOF (Local Outlier Factor) algorithm is a density-based anomaly detection method. This algorithm calculates the density of data points and their corresponding values. The local reachability density of neighboring points is used to derive the local temperature rise anomaly factor at that point. When A value significantly greater than 1 indicates that the local density of the point is lower than the average local density of its neighboring points, and therefore it is identified as a local outlier. The incremental nature of this algorithm allows it to adapt to streaming data, eliminating the need to repeatedly calculate all historical data and improving real-time performance.

[0004] However, directly applying existing LOF algorithms to motor temperature monitoring scenarios has inherent limitations: strong coupling and high noise: Motor temperature is strongly coupled with motor load current and ambient temperature. Normal load fluctuations or ambient temperature changes are sufficient to cause the motor temperature to deviate from its historical average, forming sparse normal operating condition transition regions in the feature space. In this case, the standard LOF algorithm struggles to distinguish between this normal local density transition and subtle anomalies caused by initial faults, leading to a higher false alarm rate. Insensitivity to evolving anomalies: Initial motor faults are usually a slow evolving process, manifested as a sustained temperature rise above the expected value over a period of time, rather than a sudden spike. This slow, continuous deviation may not be identified as a significant anomaly in the static local density calculation of the standard LOF algorithm, leading to a higher false negative rate. Summary of the Invention This invention provides a fault early warning method and system based on motor temperature monitoring, aiming to solve the problem that the standard LOF algorithm in related technologies has difficulty distinguishing between normal local density shifts and subtle anomalies caused by initial faults, leading to an increased false alarm rate.

[0005] In a first aspect, the present invention provides a fault early warning method based on motor temperature monitoring, comprising: acquiring operating data including motor winding temperature, ambient temperature, and effective value of load current; calculating a temperature rise anomaly factor, wherein the temperature rise anomaly factor is positively correlated with the difference between the motor winding temperature and the ambient temperature, and negatively correlated with the product of the square of the effective value of the load current and a preset motor thermal resistance; taking the difference between the temperature rise anomaly factor at each time point and its corresponding weighted moving average as the instantaneous temperature rise deviation; and for each data point, determining its... Nearest neighbor set, calculate the The variance of the instantaneous temperature rise deviation corresponding to all data points within the nearest neighbor set is used as the local fluctuation stability of that data point; a local density fluctuation factor is obtained based on the local fluctuation stability, and the value of the local density fluctuation factor is negatively correlated with the local fluctuation stability; the data point is then mapped to its nearest neighbor set. The reachability distance of each neighbor point in the nearest neighbor set is divided by the local density fluctuation factor to obtain the corrected reachability distance; the corrected local reachability density of the data point is calculated based on the corrected reachability distance; the anomaly score of the data point is calculated, and the anomaly score is... The ratio of the average corrected local reachability density of the nearest neighbor set to the corrected local reachability density of the data point is used to trigger a fault warning based on the anomaly score of each data point. First, a temperature rise anomaly factor is constructed based on the thermal balance principle to physically isolate load and environmental interference. Second, a local density fluctuation factor is introduced at the data level, using the fluctuation variance of the data point's neighborhood to adaptively correct the distance metric in the LOF algorithm. This mechanism creatively transforms the instability of the initial stage of a fault into low-density features in the feature space, solving the problem of traditional methods' difficulty in distinguishing between normal operating condition migration and slowly evolving faults, significantly improving the detection accuracy of early-stage motor faults.

[0006] Furthermore, the temperature rise anomaly factor is calculated using the following formula: In the formula, Indicates in Temperature rise anomaly factor at any given time; for The temperature of the motor windings is measured at all times; for The ambient temperature measured at all times; for The effective value of the motor load current measured at constant time, taken as square here. To correspond to the Joule heating effect; This represents the equivalent thermal resistance coefficient of the motor. Based on the principle of motor thermodynamic balance, a specific quantitative formula for the temperature rise anomaly factor is provided. This formula can accurately subtract the normal temperature rise component caused by the Joule heating effect (proportional to the square of the current) and the ambient reference temperature from the original monitored temperature, thereby obtaining the core characteristic indicators that are only related to the internal insulation or mechanical health status of the motor, providing a clean data foundation for subsequent anomaly detection.

[0007] Furthermore, the weighted moving average is calculated using an exponentially weighted moving average algorithm. This algorithm can effectively simulate the large thermal inertia characteristics of the motor itself, establish a dynamic benchmark through smoothing historical data, and help filter out interference from high-frequency random noise, thereby more accurately capturing the instantaneous temperature rise deviation that breaks the thermal equilibrium through the difference.

[0008] Furthermore, the formula for calculating the weighted moving average is: In the formula, for The weighted moving average of the temperature rise anomaly factor over time. For smoothing coefficients, for Temperature rise anomaly factor at any given moment. By introducing a smoothing coefficient, the system can appropriately respond to changes at the current moment while retaining the memory of historical thermal states, ensuring the dynamic tracking capability of the reference value, so that the calculated instantaneous temperature rise deviation can truly reflect the sudden changes in the motor's thermal state.

[0009] Furthermore, the formula for calculating the local density fluctuation factor is as follows: In the formula, For data points Local density fluctuation factor; For data points Local fluctuation stability; The sensitivity coefficient, It is an exponential function with the natural constant e as its base. Through the exponential function relationship, when the data in the region fluctuates greatly (i.e., the stability is poor, corresponding to the potential risk of failure), the value of this factor will decrease significantly. In this way, the equivalent reachable distance of the point will be amplified in subsequent calculations, and its local density will be reduced, making unstable fault precursor data easier to identify as outliers in the feature space.

[0010] Furthermore, the range of the sensitivity coefficient is as follows: .

[0011] Furthermore, the method for calculating the corrected local reachability density is as follows: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The number of nearest neighbors is divided by the sum of the corrected reachable distances to all neighboring points. This clarifies the calculation path for the corrected local reachability density, ensuring that the fluctuation factor introduced in the previous steps is correctly integrated into the density calculation of the LOF algorithm. This allows the density value to not only reflect the spatial density but also incorporate the stability information of the data distribution, which is a key step in achieving high-sensitivity anomaly detection.

[0012] Furthermore, the process includes time alignment and normalization preprocessing of the collected winding temperature, ambient temperature, and load current data. This eliminates the asynchrony in time sampling and the huge differences in dimensions and orders of magnitude of multi-source sensor data (temperature, current), preventing a single feature value from dominating distance calculations and ensuring the fairness and accuracy of multi-dimensional feature data in Euclidean space distance calculations.

[0013] Furthermore, based on the magnitude of the abnormal scores for each data point, a fault warning is triggered, including: in response to an abnormal score for a data point exceeding an abnormal threshold, a fault warning is triggered, wherein the abnormal threshold is 1.2 to 1.5. This converts continuously changing abnormal scores into discrete alarm signals, achieving automated monitoring of the motor's operating status.

[0014] In a second aspect, a fault early warning system based on motor temperature monitoring includes a processor and a memory, the memory storing a computer program, the processor executing the computer program to implement the fault early warning method based on motor temperature monitoring as described in any of the preceding claims.

[0015] Beneficial effects: By constructing a temperature rise anomaly factor and combining it with instantaneous temperature rise deviation, the strong coupling effect of load current fluctuations and ambient temperature changes on motor winding temperature is effectively isolated, solving the problem that existing technologies easily misjudge temperature migration under normal operating conditions as faults. At the same time, the introduction of a local density fluctuation factor adaptively corrects the distance metric in the traditional LOF (Local Anomaly Factor) algorithm, using the fluctuation stability (variance) within the neighborhood of data points to adjust the local density, so that data points that are in the fault evolution period, although not significantly deviating but exhibiting an unstable state, are assigned lower density values, thereby significantly improving the algorithm's sensitivity to slowly evolving initial faults in motors and reducing the false alarm rate. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating a temperature monitoring flowchart according to an embodiment of the present invention. Detailed Implementation

[0017] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] like Figure 1 As shown, S101: Multi-source data acquisition and preprocessing.

[0019] In this embodiment, the system first collects multi-source data in real time during motor operation using a sensor array deployed at key parts of the motor. Specifically, the winding temperature of the motor is collected using a platinum resistance thermometer embedded in the stator slot, the ambient temperature is collected using an ambient temperature sensor installed around the motor, and the three-phase load current of the motor is collected using a Hall current sensor and its effective value is calculated.

[0020] After acquiring the raw data, it undergoes time alignment and cleaning. Since different sensors may have different sampling frequencies, linear interpolation is used to unify all data to the same timestamp. Furthermore, the data is normalized to eliminate the impact of dimensional differences on subsequent distance calculations.

[0021] S102: Calculate the temperature rise anomaly factor at each time point.

[0022] During motor operation, its temperature changes depend not only on its internal health condition but also on the external load and ambient temperature. To eliminate interference from external factors, an abnormal temperature rise factor needs to be constructed.

[0023] In this embodiment, based on the principle of motor thermodynamic balance, the motor's temperature rise mainly originates from copper losses (proportional to the square of the current) and is constrained by heat dissipation conditions (ambient temperature). Therefore, the temperature rise anomaly factor can be characterized by constructing the difference between the actual temperature rise and the theoretical temperature rise. The calculation formula is as follows: In the formula, Indicates in Temperature rise anomaly factor at any given time; for The temperature of the motor windings is measured at all times; for The ambient temperature measured at all times; for The effective value of the motor load current measured at constant time, taken as square here. To correspond to the Joule heating effect; This is the equivalent thermal resistance coefficient of the motor.

[0024] In this embodiment, It is a pre-calibrated constant. As a preferred option, The acquisition method is as follows: under different load conditions, multiple sets of data are collected when the motor is in its factory healthy state. The data was obtained by linear regression fitting using the least squares method. For example, after calibration... for .

[0025] As can be seen from the above formula, when the motor is in a healthy state, the actual temperature rise is... Should be in line with theoretical temperature rise Basically the same, The value in Fluctuations in the vicinity. If an initial fault occurs inside the motor, such as bearing wear or inter-turn short circuit, it will generate additional internal heat, leading to... An abnormal increase occurred, while the load at this time... and environment Unchanged, making The value increases significantly. This step effectively separates the coupling effects of the external environment and load from the raw temperature data.

[0026] S103: Calculate the instantaneous temperature rise deviation at each moment.

[0027] Considering the significant thermal inertia of motors, temperature changes caused by initial faults typically exhibit a slow evolution rather than an instantaneous spike. To capture this evolutionary characteristic and filter out high-frequency noise, the instantaneous temperature rise deviation needs to be calculated. First, the long-term trend term of the temperature rise anomaly factor is calculated using the Exponentially Weighted Moving Average (EWMA) algorithm. The calculation formula is as follows: Next, calculate the instantaneous temperature rise deviation using the following formula: In the formula, for Instantaneous temperature rise deviation at any given moment; for The weighted moving average of the temperature rise anomaly factor over time; This is the smoothing coefficient. Preferably, it is... Set as .

[0028] In the formula, This represents the expected stable state after taking thermal inertia into account. When the motor is running smoothly, Its smoothing value Very close. Approaching When a fault causes an abnormal temperature rise factor When deviation begins to occur, due to It has a lag effect. It will quickly pull apart from The gap led to The absolute value increases. Therefore, It can keenly detect the early signs of a disruption in thermal equilibrium.

[0029] S104: Calculate the local density fluctuation factor and adaptively adjust the local reachable density.

[0030] Traditional Local Anomaly Factor (LOF) algorithms rely solely on static spatial distance, making it difficult to distinguish between sparse regions under normal operating conditions and fluctuating regions caused by faults. This step introduces a local density fluctuation factor to correct the distance metric, making the algorithm sensitive to instability.

[0031] First, for the current data point In its Nearest neighbor set Within the set, calculate the instantaneous temperature rise deviation corresponding to all data points. The variance of the local fluctuation stability is denoted as the variance of the local fluctuation stability. Subsequently, a local density fluctuation factor was constructed. The calculation formula is as follows: In the formula, For data points Local density fluctuation factor; For data points Local fluctuation stability (i.e., within the neighborhood) (variance) This refers to the sensitivity coefficient. In this embodiment, the sensitivity coefficient... The preset positive number is preferred, and the range is: For example, setting .

[0032] As can be seen from the formula, when the thermal state of the motor in a certain area is very stable, the instantaneous temperature rise deviation is... The fluctuations are minimal. Approaching ,at this time Approaching Conversely, if a region is in the fault evolution phase and its thermal state is unstable, Enlargement, leading to It decays exponentially, becoming less than The value.

[0033] Finally, this factor is used to correct the local reachability density in the LOF algorithm. The corrected calculation formula is as follows: In the formula, For data points The corrected local achievable density; For data points To its neighbor's point The original reachable distance is usually the Euclidean distance; for of Nearest neighbor set; For the number of neighbors, for example, take .

[0034] In this formula, Located in the denominator and used as distance The divisor. When the monitoring point is in the initial stage of a fault with an unstable state, due to... Larger It will become smaller (e.g.) At this point, the corrected equivalent distance term It will increase. This is equivalent to increasing the distance between the point and its neighbors in the feature space. As the formula shows, increasing the denominator (the sum of distances) will lead to an increase in the final calculated local reachability density. Significantly reduced. In other words, this mechanism treats data points with large thermal state fluctuations as low-density points, making them easier to identify as anomalies in subsequent LOF calculations, thus solving the problem of missed detections caused by evolutionary faults not being physically far enough from normal clusters.

[0035] S105: Calculate the anomaly score and issue a fault warning.

[0036] The final anomaly score is calculated based on the corrected local reachability density. : In the formula, For data points Abnormal scores; Point for neighbors The corrected locally attainable density; For data points The corrected local achievable density.

[0037] If data points It's normal; its density... Density with its neighbors Similar, ratios are close If data points There is a risk of failure, as can be seen from step S104, its correction density It has been significantly suppressed, while the density of its neighboring points (which may be historical normal data) has been significantly reduced. A higher value results in a larger numerator and a smaller denominator, leading to... Significantly greater than .

[0038] In this embodiment, the abnormal threshold is set to That is: if The system determines that the motor has an initial fault risk and immediately triggers an audible and visual alarm or sends a stop command to the host computer; if The motor is determined to be operating normally, and the current data point is recorded. Update to the historical database for subsequent incremental calculations.

[0039] The present invention also provides a fault early warning system based on motor temperature monitoring. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the fault early warning method based on motor temperature monitoring according to the first aspect of the present invention.

[0040] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0041] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0042] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A fault early warning method based on motor temperature monitoring, characterized in that, include: Acquire operating data including motor winding temperature, ambient temperature, and RMS load current; Calculate the temperature rise anomaly factor, which is positively correlated with the difference between the motor winding temperature and the ambient temperature, and negatively correlated with the product of the square of the effective value of the load current and the preset motor thermal resistance; the difference between the temperature rise anomaly factor at each moment and its corresponding weighted moving average is taken as the instantaneous temperature rise deviation. For each data point, determine its Nearest neighbor set, calculate the The variance of the instantaneous temperature rise deviation corresponding to all data points within the nearest neighbor set is used as the local fluctuation stability of that data point; a local density fluctuation factor is obtained based on the local fluctuation stability, and the value of the local density fluctuation factor is negatively correlated with the local fluctuation stability; the data point is then mapped to its nearest neighbor set. The corrected reachability distance is obtained by dividing the reachability distance of each neighbor point in the nearest neighbor set by the local density fluctuation factor; the corrected local reachability density of the data point is then calculated based on the corrected reachability distance. Calculate the anomaly score for the data point, where the anomaly score is... The ratio of the average corrected local reachability density of the nearest neighbor set to the corrected local reachability density of the data point; triggering a fault warning based on the anomaly score of each data point.

2. The fault early warning method based on motor temperature monitoring according to claim 1, characterized in that, The temperature rise anomaly factor is calculated using the following formula: ； In the formula, Indicates in Temperature rise anomaly factor at any given time; for The temperature of the motor windings is measured at all times; for The ambient temperature measured at all times; for The effective value of the motor load current measured at constant time, taken as square here. To correspond to the Joule heating effect; This is the equivalent thermal resistance coefficient of the motor.

3. The fault early warning method based on motor temperature monitoring according to claim 1, characterized in that, The weighted moving average is calculated using an exponentially weighted moving average algorithm.

4. The fault early warning method based on motor temperature monitoring according to claim 3, characterized in that, The formula for calculating the weighted moving average is: ； In the formula, for The weighted moving average of the temperature rise anomaly factor over time. For smoothing coefficients, for Temperature rise anomaly factor at any given time.

5. The fault early warning method based on motor temperature monitoring according to claim 1, characterized in that, The formula for calculating the local density fluctuation factor is as follows: ； In the formula, For data points Local density fluctuation factor; For data points Local fluctuation stability; The sensitivity coefficient, It is an exponential function with the natural constant e as its base.

6. The fault early warning method based on motor temperature monitoring according to claim 5, characterized in that, The sensitivity coefficient takes values ​​ranging from 1 to 10. .

7. The fault early warning method based on motor temperature monitoring according to claim 1, characterized in that, The method for calculating the corrected local reachability density is as follows: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The number of nearest neighbors divided by the sum of the distances to all neighboring points can be obtained.

8. The fault early warning method based on motor temperature monitoring according to claim 1, characterized in that, It also includes time alignment and normalization preprocessing of the collected winding temperature, ambient temperature and load current data.

9. The fault early warning method based on motor temperature monitoring according to claim 1, characterized in that, Based on the magnitude of the abnormal scores for each data point, a fault warning is triggered, including: If the abnormal score of a data point is greater than the abnormal threshold, a fault warning is triggered, wherein the abnormal threshold is 1.2 to 1.

5.

10. A fault early warning system based on motor temperature monitoring, comprising a processor and a memory, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the fault early warning method based on motor temperature monitoring as described in any one of claims 1-9.

Images