A power equipment fault detection system based on a large model

Abstract

This invention relates to the field of fault detection technology, and more particularly to a power equipment fault detection system based on a large model. The system includes a data acquisition module, a data analysis module, a latent variable extraction module, a fault early warning module, and a threshold adjustment module. This invention eliminates the interference of coupled component temperatures (such as gearboxes and frequency converters) on generator bearing temperatures through the latent variable extraction module, accurately distinguishing between actual faults and coupled interference. By identifying and processing the influence of coupled component temperatures on bearing temperatures, the accuracy of detection is improved. Simultaneous analysis of three types of features—temperature gradient, vibration characteristics, and current harmonics—enhances the detection rate of early micro-corrosion pits, providing earlier warnings. Based on the latent variable influence model and by calculating the time difference between alarm time and fault time, a warning margin is obtained, enabling precise early warning of electrical corrosion faults in power equipment.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology, and in particular to a power equipment fault detection system based on a large model. Background Technology

[0002] During operation, power equipment operates in a complex environment. Humidity, air containing corrosive substances, and uneven current distribution can all lead to micro-current corrosion, thereby reducing the operating efficiency of power equipment and even causing malfunctions. Timely detection and elimination of potential faults in power equipment are of great significance for reducing power loss caused by downtime due to malfunctions and ensuring the stability and reliability of its operation.

[0003] Chinese Patent Publication No. CN117171709B discloses a method and system for fault detection of permanent magnet generators. The method includes: obtaining a predetermined detection cycle of the target permanent magnet generator; collecting feature values ​​to obtain a set of Q feature values ​​for state monitoring indicators; obtaining deviation coefficients of the Q state monitoring indicators; obtaining Q state detection nodes; performing fusion analysis to obtain a set of target fault detection nodes; performing union on multiple state-related detection items corresponding to the set of target fault detection nodes to obtain a set of multiple target-related detection items; and performing fault detection on the target permanent magnet generator based on the set of target fault detection nodes and the set of multiple target-related detection items. It is evident that existing fault detection processes are susceptible to interference from coupling components, leading to inaccurate detection of micro-current corrosion faults and resulting in low operating efficiency of the power generation equipment. Summary of the Invention

[0004] To address this issue, the present invention provides a power equipment fault detection system based on a large model, which overcomes the problem in the prior art of lacking the ability to identify the influence of the temperature of coupling components on the bearing temperature, leading to misjudgment of faults and low operating efficiency of power generation equipment.

[0005] To achieve the above objectives, the present invention provides a power equipment fault detection system based on a large model, comprising:

[0006] The data acquisition module is connected to several sensors installed inside the power equipment to collect the temperature difference and shaft current characteristics at both ends of the generator bearing.

[0007] The data analysis module, which is connected to the data acquisition module, is used to calculate the axial temperature gradient of the bearing based on the temperature difference between the two ends of the axial direction, determine the operating state of the bearing based on the comparison between the axial temperature gradient and the temperature gradient threshold, and analyze the area of ​​the overheated region of the bearing in combination with the shaft current characteristics and the bearing wear condition to obtain the analysis results.

[0008] The latent variable extraction module, which is connected to the data analysis module, is used to extract the interference characteristics of the coupling components based on the analysis results and to perform temperature interference processing on the coupling components.

[0009] A fault early warning module, which is connected to the data analysis module, is used to issue corresponding early warnings based on the analysis results;

[0010] The threshold adjustment module, which is connected to the data analysis module, is used to calculate the warning margin for electrical corrosion fault warning and determine whether to adjust the overheated area threshold based on the comparison between the warning margin and the margin threshold range.

[0011] Furthermore, the data analysis module includes an overheated area analysis unit, a bearing wear analysis unit, and a shaft current analysis unit, wherein:

[0012] The overheated area analysis unit is used to identify local overheated areas of the bearing based on the comparison results of the bearing axial temperature gradient and temperature gradient threshold, and calculate the real-time overheated area. The real-time overheated area is compared with the standard overheated area to obtain the first comparison result and the second comparison result.

[0013] The bearing wear analysis unit is used to call historical information from the database for correlation analysis when the first comparison result is obtained, so as to output wear-related features.

[0014] The shaft current analysis unit is used to analyze the shaft current characteristics when the second comparison result is obtained. The shaft current characteristics include shaft current amplitude and distortion rate.

[0015] Furthermore, the overheated area analysis unit includes a temperature gradient calculation subunit, a first comparative analysis subunit, and a second comparative analysis subunit, wherein:

[0016] The temperature gradient calculation subunit is used to calculate the axial temperature gradient of the bearing based on the temperature difference between the two ends of the axial direction and the bearing length.

[0017] The first comparative analysis subunit is used to compare the bearing axial temperature gradient with the temperature gradient threshold, and determine whether there is any abnormality in the bearing's operating state based on the comparison results;

[0018] The second comparative analysis subunit is used to compare the real-time superheated area with the standard superheated area to obtain the first comparison result and the second comparison result.

[0019] Furthermore, the first comparative analysis subunit determines that the bearing's operating state is abnormal when the bearing's axial temperature gradient is greater than the temperature gradient threshold.

[0020] When the axial temperature gradient of the bearing is less than or equal to the temperature gradient threshold, the bearing is considered to be in normal operating condition.

[0021] Furthermore, the data analysis module also includes a corrosion product analysis unit;

[0022] The corrosion product analysis unit is used to determine the degree of corrosion product accumulation based on the hot spot shape and edge temperature gradient.

[0023] Furthermore, the fault early warning module includes a stress concentration early warning unit and an electro-corrosion fault early warning unit, wherein,

[0024] The stress concentration early warning unit is used to trigger a stress concentration early warning prompt based on the degree of corrosion product accumulation.

[0025] The electro-corrosion fault early warning unit is used to trigger an electro-corrosion fault early warning based on the bearing wear condition and the duration of temperature, as well as based on the shaft current characteristics.

[0026] The lubrication failure early warning unit is used to trigger lubrication failure early warning prompts based on shaft current characteristics, the degree of corrosion product accumulation, the duration of temperature change, and the rate of temperature change.

[0027] Furthermore, the latent variable extraction module includes a detection unit, an extraction unit, and an execution unit, wherein:

[0028] The detection unit is used to detect the temperature of the coupling component based on the degree of corrosion product accumulation and the duration of temperature.

[0029] The extraction unit is used to extract interference features of the coupling components based on the detection results;

[0030] The execution unit is used to perform temperature interference processing of the coupling component according to the interference characteristics of the coupling component.

[0031] Furthermore, the electro-corrosion fault early warning unit includes a first electro-corrosion fault early warning subunit and a second electro-corrosion fault early warning subunit.

[0032] The first electro-corrosion fault early warning subunit triggers an electro-corrosion fault early warning prompt based on the shaft current characteristics.

[0033] The second electro-corrosion fault early warning subunit triggers an electro-corrosion fault early warning prompt when it determines that there is no wear and the temperature duration exceeds the duration threshold.

[0034] Furthermore, the bearing wear analysis unit performs correlation analysis by calling historical information from the database, including:

[0035] Collect vibration signals and perform preprocessing;

[0036] After preprocessing, the sample is determined to be worn or not worn based on kurtosis and sample entropy.

[0037] Furthermore, the threshold adjustment module includes a margin calculation unit, a comparison unit, and an adjustment unit, wherein:

[0038] The margin calculation unit is used to calculate the time difference between the alarm time and the fault time to obtain the warning margin;

[0039] The comparison unit is used to compare the warning margin with the margin threshold range;

[0040] The adjustment unit is used to determine whether to adjust the overheated area threshold based on the comparison results.

[0041] Compared with existing technologies, the beneficial effects of this invention are as follows: by eliminating the interference of the temperature of coupled components such as gearbox and frequency converter on the temperature of generator bearing through the latent variable extraction module, the true fault and coupled interference can be accurately distinguished. By identifying and processing the influence of coupled component temperature on bearing temperature, the accuracy of detection is improved. By simultaneously analyzing three types of features—temperature gradient, vibration characteristics, and current harmonics—the detection rate of early micro-corrosion pits is improved, and the early warning time is advanced. Based on the latent variable influence model and by calculating the time difference between alarm time and fault time, the early warning margin is obtained, and accurate early warning is achieved, realizing accurate early warning of electrical corrosion faults in power equipment. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the structure of a power equipment fault detection system based on a large model according to an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of the data analysis module in an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the latent variable extraction module according to an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of the threshold adjustment module in an embodiment of the present invention. Detailed Implementation

[0046] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0047] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0048] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0049] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0050] Please see Figure 1 The diagram shown is a structural schematic of a power equipment fault detection system based on a large model, according to an embodiment of the present invention. The present invention provides a power equipment fault detection system based on a large model, comprising:

[0051] The data acquisition module is connected to several sensors installed inside the power equipment to collect the temperature difference and shaft current characteristics at both ends of the generator bearing.

[0052] The data analysis module, which is connected to the data acquisition module, is used to calculate the axial temperature gradient of the bearing based on the temperature difference between the two ends of the axial direction, determine the operating state of the bearing based on the comparison between the axial temperature gradient and the temperature gradient threshold, and analyze the area of ​​the overheated region of the bearing in combination with the shaft current characteristics and the bearing wear condition to obtain the analysis results.

[0053] The latent variable extraction module, which is connected to the data analysis module, is used to extract the interference characteristics of the coupling components based on the analysis results and to perform temperature interference processing on the coupling components.

[0054] A fault early warning module, which is connected to the data analysis module, is used to issue corresponding early warnings based on the analysis results;

[0055] The threshold adjustment module, which is connected to the data analysis module, is used to calculate the warning margin for electrical corrosion fault warning and determine whether to adjust the overheated area threshold based on the comparison between the warning margin and the margin threshold range.

[0056] In this embodiment, the power equipment is a permanent magnet synchronous generator. Since the electro-corrosion of permanent magnet synchronous generators is mainly caused by shaft current, when the current is transmitted from one end to the other, it will break down the oil film on the bearing, thereby generating shaft current and causing electrochemical corrosion. Long-term electro-corrosion can easily cause the formation of washboard marks and bearing failure, leading to malfunctions. Therefore, by monitoring the overheating of the bearing and analyzing the causes of overheating, the interference of the temperature of coupling components such as gearbox and frequency converter on the temperature of generator bearing is eliminated through the latent variable extraction module, accurately distinguishing between real faults and coupling interference. By identifying and processing the influence of the temperature of coupling components on the bearing temperature, the accuracy of detection is improved. At the same time, the detection rate of early micro-corrosion pits is improved by analyzing three types of characteristics: temperature gradient, vibration characteristics, and current harmonics. Based on the latent variable influence model and by calculating the time difference between alarm time and fault time, the warning margin is calculated. The model parameters are optimized based on the warning margin to improve the timeliness of the warning and ensure accurate early warning of electro-corrosion faults of power equipment. That is, timely detection of potential faults in power equipment, reducing power generation losses caused by fault downtime, and ensuring the stability and reliability of power equipment operation.

[0057] See Figure 2 As shown, it is a structural schematic diagram of the data analysis module in an embodiment of the present invention;

[0058] Specifically, the data analysis module includes an overheated area analysis unit, a bearing wear analysis unit, and a shaft current analysis unit, wherein:

[0059] The overheated area analysis unit is used to identify local overheated areas of the bearing based on the comparison results of the bearing axial temperature gradient and temperature gradient threshold, and calculate the real-time overheated area. The real-time overheated area is compared with the standard overheated area to obtain the first comparison result and the second comparison result.

[0060] The bearing wear analysis unit is used to call historical information from the database for correlation analysis when the first comparison result is obtained, so as to output wear-related features.

[0061] The shaft current analysis unit is used to analyze the shaft current characteristics when the second comparison result is obtained. The shaft current characteristics include shaft current amplitude and distortion rate.

[0062] Specifically, the overheated area analysis unit includes a temperature gradient calculation subunit, a first comparative analysis subunit, and a second comparative analysis subunit, wherein:

[0063] The temperature gradient calculation subunit is used to calculate the axial temperature gradient of the bearing based on the temperature difference between the two ends of the axial direction and the bearing length.

[0064] The first comparative analysis subunit is used to compare the bearing axial temperature gradient with the temperature gradient threshold, and determine whether there is any abnormality in the bearing's operating state based on the comparison results;

[0065] The second comparative analysis subunit is used to compare the real-time superheated area with the standard superheated area to obtain the first comparison result and the second comparison result;

[0066] Among them, when the real-time superheated area is smaller than the standard superheated area, the first comparison result is obtained; when the real-time superheated area is greater than or equal to the standard superheated area, the second comparison result is obtained.

[0067] Specifically, the first comparative analysis subunit determines that the bearing's operating state is abnormal when the bearing's axial temperature gradient is greater than the temperature gradient threshold.

[0068] When the axial temperature gradient of the bearing is less than or equal to the temperature gradient threshold, the bearing is considered to be in normal operating condition.

[0069] In this embodiment, the temperature gradient is the rate of temperature change per unit length along the bearing axis, expressed in °C / m. The temperature gradient threshold is set between 5 °C / m and 15 °C / m, preferably 15 °C / m. This threshold characterizes the temperature under normal bearing operation. The temperature difference between the two axial ends is the absolute value of the temperature difference between the two ends of the bearing, specifically the difference between the measured temperatures Ts1 and Ts2. The bearing axial temperature gradient is calculated based on the temperature difference between the two ends of the bearing and the bearing length. For example, if the bearing length is 0.1m and the temperature difference between the two axial ends is 2 °C, then the bearing axial temperature gradient is 20 °C / m, which is greater than 15 °C / m. At this point, the temperature gradient is greater than the temperature gradient threshold. When the temperature gradient is determined to be greater than the temperature gradient threshold, the bearing operation is abnormal. The local overheated area of ​​the bearing is identified and the real-time overheated area is calculated to accurately identify the cause of the abnormality. The real-time overheated area is the sum of the areas of the overheated areas of the bearing. When the temperature gradient is determined to be less than or equal to the temperature gradient threshold, the bearing operation is normal. The standard overheated area is a threshold for distinguishing the size of the overheated area, set at 5% of the bearing surface area. When the real-time overheated area is greater than the standard overheated area, a second comparison result is obtained. When the real-time overheated area is less than or equal to the standard overheated area, a first comparison result is obtained.

[0070] Because the heat generated by the generator's gearbox or frequency converter during operation can be conducted to the bearings through the mechanically connected spindle or thermal radiation, the bearing temperature may appear artificially high. This masks the subtle temperature rise caused by actual electro-corrosion, making it difficult to provide early warning of electro-corrosion faults. Therefore, it is crucial to deeply correlate temperature data with physical phenomena to accurately distinguish fault types. Specifically, the bearing's operating condition can be preliminarily determined based on the temperature gradient. If the temperature gradient exceeds a certain threshold, an abnormality is considered. The cause needs to be investigated based on the size of the overheated area. Comparing the real-time overheated area with a standard overheated area indicates a small-area overheating. In this case, it could be caused by micro-corrosion pits formed by early electro-corrosion, stress concentration due to corrosion product accumulation, or minor wear. A second comparison indicates a large-area overheating, which could be caused by thermal interference from coupling components leading to an overall temperature reading deviation. Because electro-corrosion faults occur over a wide range—namely, due to oil film rupture leading to lubrication failure and resulting in heat generation from metal friction—further precise analysis is required. Specifically, upon obtaining the first comparison result, the bearing wear condition is analyzed. Historical information from the database, specifically historical vibration characteristics, is used for correlation analysis. Wear is determined using both vibration kurtosis and sample entropy as dual indicators, and combined with the duration of abnormal temperature gradients, effectively distinguishing between wear and early-stage electro-corrosion faults. When bearing wear is determined, a wear warning is issued. When no bearing wear is determined, the duration of abnormal temperature gradients is obtained. If it exceeds a threshold, it is determined to be an electro-corrosion fault; if it does not exceed the threshold, the cause of the abnormal temperature gradient is determined to be the interference of coupling component temperature on the generator bearing temperature. Upon obtaining the second comparison result, shaft current and temperature data are monitored in real time for timely warnings and fault analysis. Specifically, the shaft current amplitude of the bearing is analyzed to determine if the current is abnormal, thereby determining the presence and severity of electro-corrosion. Early detection of abnormalities in the early stages of electro-corrosion provides early warnings and prevents serious damage.

[0071] In this embodiment, the shaft current characteristic is the shaft current amplitude, which is acquired by a high-frequency current transformer at a sampling frequency of 5kHz and an accuracy of ±0.5%. The shaft current amplitude is the peak current, which is the maximum absolute value of the current signal and reflects the instantaneous impact intensity. The shaft current amplitude threshold is set to 0.5A. When the shaft current amplitude is determined to be greater than the shaft current amplitude threshold, an electro-corrosion fault is identified, triggering an electro-corrosion fault warning. At this time, due to the oil film breakdown threshold, electrical spark erosion forms pits, resulting in an electro-corrosion fault. When the shaft current amplitude is less than or equal to the threshold, the harmonic distortion rate is obtained. When the distortion rate is greater than the threshold, an electro-corrosion fault is identified, and an electro-corrosion fault warning is triggered. When the distortion rate is less than or equal to the threshold, the proportion of the third harmonic is obtained. If the proportion of the third harmonic is greater than 40%, it is identified as early electro-corrosion, and the corrosion products are analyzed. The stress concentration warning is triggered based on the degree of corrosion product accumulation. Furthermore, the interference of coupled components, electro-corrosion faults, and lubrication failures are distinguished based on the duration of temperature change and the rate of temperature change, so as to achieve precise maintenance.

[0072] Harmonic analysis is used to obtain the harmonic distortion rate. The harmonic distortion rate is collected by a spectrum analyzer at a sampling frequency of 50Hz or 60Hz. The process of harmonic analysis involves performing a fast Fourier transform on the current signal to decompose it into frequency domain components, generating a frequency coordinate axis and a corresponding array of spectral amplitudes. Harmonic components in the current are identified through spectrum analysis. The distortion rate is calculated as (total harmonic energy / fundamental amplitude) × 100%. The total harmonic energy is the root mean square value of all harmonic components. The total harmonic energy is obtained by taking the square root of the sum of the squares of the amplitudes of all harmonic components. The total harmonic energy is actually the total effective value of the harmonic components. The specific calculation process is as follows: first, the fundamental component is removed from the entire spectrum. The fundamental component is the frequency point corresponding to the maximum amplitude of the spectrum. Then, the amplitudes of all remaining frequency components are squared and summed. Finally, the square root of the sum is taken to obtain the total harmonic energy.

[0073] Specifically, the fault early warning module includes a stress concentration early warning unit and an electro-corrosion fault early warning unit, wherein,

[0074] The stress concentration early warning unit is used to trigger a stress concentration early warning prompt based on the degree of corrosion product accumulation.

[0075] The electro-corrosion fault early warning unit is used to trigger an electro-corrosion fault early warning based on the bearing wear condition and the duration of temperature, as well as based on the shaft current characteristics.

[0076] The lubrication failure early warning unit is used to trigger lubrication failure early warning prompts based on shaft current characteristics, the degree of corrosion product accumulation, the duration of temperature change, and the rate of temperature change.

[0077] The data analysis module in this embodiment also includes a corrosion product analysis unit. This unit determines the degree of corrosion product accumulation based on the hotspot shape and edge temperature gradient. It identifies the presence of radial hotspots through morphological recognition. If radial hotspots are present, it analyzes whether the edge gradient of the hotspot is greater than 25℃ / m. If the edge gradient is greater than or equal to 25℃ / m, stress concentration is detected, triggering a stress concentration warning. If the edge gradient is less than 25℃ / m, no stress concentration is detected, indicating large-area uniform overheating. The unit then obtains the duration of the temperature reading. If the duration is ≤2 hours, interference from coupled components is detected. If the duration is >2 hours, the unit analyzes the temperature change rate. If the rate is >2℃ / 10min, the temperature rises rapidly, indicating early electro-corrosion accelerating the temperature change, triggering an electro-corrosion fault warning. If the rate is ≤2℃ / 10min, the temperature change is gradual. In this case, the heat generated by electro-corrosion accelerates the oxidation, decomposition, or volatilization of the lubricant, reducing lubrication performance and causing lubrication failure, requiring a lubrication failure warning.

[0078] Specifically, the morphological recognition process involves preprocessing the acquired infrared image, including denoising and contrast enhancement; using edge detection or region growing algorithms to segment high-temperature regions from the preprocessed infrared image as hotspots; performing morphological analysis on the segmented hotspot regions to determine whether their shape is radial; the radial characteristic is that there is a central region, followed by multiple linear high-temperature regions radiating outwards; alternatively, Hough transform can be used to detect straight lines, and if multiple straight lines emanating from the same central region are detected, it is determined to be a radial hotspot.

[0079] Specifically, the calculation process of the edge gradient of the radial hotspot is as follows: In the infrared image, the edge of the radial hotspot, that is, the boundary between the high temperature area and the normal area, is determined. Along the tangent direction of the edge, the temperature change per unit length is calculated to obtain the edge gradient. That is, by taking multiple detection points at the edge, for each detection point, one point is taken along the normal direction on the high temperature side and one point is taken on the low temperature side, and the temperature difference and distance difference between the two points are calculated to obtain the gradient of the detection point. Then, the maximum value of all detection points is taken as the edge gradient of the radial hotspot.

[0080] Among them, the edge gradient is the rate of temperature change perpendicular to the edge direction, and the edge direction is the direction from high temperature to low temperature. The pixel distance in the image needs to be converted into the actual distance. The actual distance corresponding to each pixel in the infrared image needs to be measured in advance, which is related to the resolution of the infrared camera and the shooting distance.

[0081] See Figure 3 As shown, it is a structural schematic diagram of the latent variable extraction module in an embodiment of the present invention;

[0082] Specifically, the latent variable extraction module includes a detection unit, an extraction unit, and an execution unit, wherein,

[0083] The detection unit is used to detect the temperature of the coupling component based on the degree of corrosion product accumulation and the duration of temperature.

[0084] The extraction unit is used to extract interference features of the coupling components based on the detection results;

[0085] The execution unit is used to perform temperature interference processing of the coupling component according to the interference characteristics of the coupling component.

[0086] In this embodiment, the duration of temperature variation is the duration of abnormal temperature gradient. When it is determined that the bearing is not worn, the timing starts from the first time the temperature gradient exceeds 15℃ / m to obtain the duration of abnormal temperature gradient. The duration threshold is set to 2 hours. If the duration of temperature variation is >2 hours, it is determined to be an electro-corrosion fault and an electro-corrosion fault warning is triggered. If the duration of temperature variation is ≤2 hours, it is determined that there is interference from the coupling component.

[0087] When interference from coupling components is detected, the temperature of the coupling components is measured to distinguish between gearbox thermal interference and inverter thermal interference, and temperature interference is processed according to the compensation model.

[0088] Specifically, the bearing temperature is calculated based on a compensation model to handle temperature interference. A sudden temperature change in the gearbox exceeding 10°C is identified as gearbox thermal interference, and a sudden temperature change in the inverter exceeding 15°C is identified as inverter thermal interference. The compensation model is as follows: =Ts-Tm, The difference between the actual and predicted temperature is represented by Tm+. Ts is the actual measured bearing temperature, and Tm is the bearing temperature predicted by the model. The model for the predicted bearing temperature is: Tm = a × load + b × speed + c × lubrication temperature + C.

[0089] Where a, b, c, etc. are coefficients obtained from model training, and C is the intercept.

[0090] In this embodiment, the bearing temperature is predicted using the DKLVR model. The specific steps are as follows:

[0091] Collect historical data including multiple characteristics such as temperature, load, speed, and lubrication conditions;

[0092] The collected data is preprocessed, such as normalized and standardized.

[0093] The DKLVR model structure selects a linear regression model;

[0094] Using training data, the model is trained using algorithms such as least squares or gradient descent to obtain coefficients and intercepts;

[0095] After training, the coefficients of the linear regression model are: load coefficient a = 0.02℃ / kg, speed coefficient b = 0.005℃ / rpm, lubrication temperature coefficient c = 0.8, and intercept C = 10℃.

[0096] By substituting new input features (such as load and speed) into the trained model, the predicted bearing temperature can be obtained; for example, when the load is 1000kg, the speed is 1500rpm, and the lubrication temperature is 30℃, Tm=0.02×1000+0.005×1500+0.8×30+10=61.5℃.

[0097] The process of temperature interference processing based on the compensation model is as follows: first, the temperature at both ends of the bearing is calculated and adjusted according to the compensation model and the bearing temperature predicted by the model; then, the temperature difference between the two ends of the axial direction is calculated and adjusted; then, the overheating area analysis unit re-identifies the local overheating area of ​​the bearing and calculates the real-time overheating area based on the comparison results of the bearing axial temperature gradient and temperature gradient threshold, and performs subsequent judgment operations.

[0098] Specifically, the electro-corrosion fault early warning unit includes a first electro-corrosion fault early warning subunit and a second electro-corrosion fault early warning subunit.

[0099] The first electro-corrosion fault early warning subunit triggers an electro-corrosion fault early warning prompt based on the shaft current characteristics.

[0100] The second electro-corrosion fault early warning subunit triggers an electro-corrosion fault early warning prompt when it determines that there is no wear and the temperature duration exceeds the duration threshold.

[0101] Specifically, the bearing wear analysis unit performs correlation analysis by calling historical information from the database, including:

[0102] Collect vibration signals and perform preprocessing;

[0103] After preprocessing, the sample is determined to be worn or not worn based on kurtosis and sample entropy.

[0104] In this embodiment, a triaxial accelerometer is installed in the radial load area of ​​the bearing housing to collect the vibration signal of the bearing at a sampling frequency of 20kHz. After preprocessing the extracted vibration signal, feature extraction is performed. The preprocessing uses minimum entropy deconvolution (MED) to enhance the impact component and suppress background noise. The feature extraction process after preprocessing involves calculating the kurtosis value of the MED-processed signal, obtaining the kurtosis, and calculating the sample entropy to measure signal complexity. The sample entropy calculation process is as follows: given the embedding dimension m and tolerance r, where r = 0.2 times the standard deviation; construct a vector sequence of length m; calculate the vector pair (B) satisfying maximum distance ≤ r and the vector pair (A) of m+1 dimensions; sample entropy = -ln( (A / B) Kurtosis is used to characterize impact intensity. Normal bearing vibration kurtosis is around 3, while wear will lead to increased impact. Kurtosis threshold is used to measure the critical value of impact intensity caused by wear. The set value is affected by the load. The kurtosis threshold can be set to 4. Under high load, the kurtosis threshold can be appropriately increased. For example, when the load is >80%, the threshold can be adjusted to 4.2. Sample entropy is used to measure signal complexity. Wear will increase the periodicity of the vibration signal, which will lead to a decrease in sample entropy. Under normal non-wear conditions, sample entropy is >1.5, while under wear conditions, sample entropy is <1.2. When the kurtosis is >4 and the sample entropy is <1.2, the bearing is judged to be worn; otherwise, the bearing is judged not to be worn. Wear is judged by both vibration kurtosis and sample entropy, which improves reliability.

[0105] See Figure 4 As shown, it is a structural schematic diagram of the threshold adjustment module in an embodiment of the present invention;

[0106] Specifically, the threshold adjustment module includes a margin calculation unit, a comparison unit, and an adjustment unit, wherein:

[0107] The margin calculation unit is used to calculate the time difference between the alarm time and the fault time to obtain the warning margin;

[0108] The comparison unit is used to compare the warning margin with the margin threshold range;

[0109] The adjustment unit is used to determine whether to adjust the overheated area threshold based on the comparison results.

[0110] In this embodiment, the alarm time is the time when an electrical corrosion fault warning is issued, and the fault time is the actual time when the predicted fault (bearing lubrication failure and stress concentration) finally occurs. The minimum and maximum margins constitute the margin threshold range, which is related to the predicted fault type. When the fault is determined to be bearing lubrication failure, the minimum margin is set to 24 hours, the maximum margin is set to 72 hours, and the target margin is 48 hours. When the fault is determined to be stress concentration, the minimum margin is set to 12 hours, the maximum margin is set to 48 hours, and the target margin is 24 hours. When the warning margin is determined to be less than or equal to the minimum margin threshold, the overheated area threshold is reduced according to the corresponding target margin. The adjusted overheated area threshold = overheated area. Threshold × (1 - min{1.0%, (target margin - warning margin) / target margin}); When the warning margin is determined to be greater than or equal to the maximum margin threshold, the overheated area threshold is increased according to the corresponding target margin. The adjustment formula is: Adjusted overheated area threshold = overheated area threshold × [1 + min{0.5%, (target margin - warning margin) / target margin}]; When the warning margin is determined to be within the margin threshold range, the overheated area threshold is not adjusted; For example, when it is determined to be stress concentration, the warning margin is 10.5 hours < 12 hours, then the overheated area threshold is adjusted to 5% × [1 - min{1.0%, (24 - 10.5) / 24}] = 4.75%.

[0111] By using a closed-loop feedback mechanism to achieve adaptive optimization of the threshold, the problem of fixed thresholds being difficult to adapt to complex working conditions is solved, while improving the accuracy and timeliness of the system's early warning.

[0112] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0113] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A power equipment fault detection system based on a large model, characterized in that, include: The data acquisition module is connected to several sensors installed inside the power equipment to collect the temperature difference and shaft current characteristics at both ends of the generator bearing. The data analysis module, which is connected to the data acquisition module, is used to calculate the axial temperature gradient of the bearing based on the temperature difference between the two ends of the axial direction, determine the operating state of the bearing based on the comparison between the axial temperature gradient and the temperature gradient threshold, and analyze the area of ​​the overheated region of the bearing in combination with the shaft current characteristics and the bearing wear condition to obtain the analysis results. The latent variable extraction module, which is connected to the data analysis module, is used to extract the interference characteristics of the coupling components based on the analysis results and to perform temperature interference processing on the coupling components. A fault early warning module, which is connected to the data analysis module, is used to issue corresponding early warnings based on the analysis results; The threshold adjustment module, which is connected to the data analysis module, is used to calculate the warning margin for electrical corrosion fault warning and determine whether to adjust the overheated area threshold based on the comparison between the warning margin and the margin threshold range. The data analysis module includes an overheated area analysis unit, a bearing wear analysis unit, and a shaft current analysis unit, wherein: The overheated area analysis unit is used to identify local overheated areas of the bearing based on the comparison results of the bearing axial temperature gradient and temperature gradient threshold, and calculate the real-time overheated area. The real-time overheated area is compared with the standard overheated area to obtain the first comparison result and the second comparison result. The bearing wear analysis unit is used to call historical information from the database for correlation analysis when the first comparison result is obtained, so as to output wear-related features. The shaft current analysis unit is used to analyze the shaft current characteristics when the second comparison result is obtained. The shaft current characteristics include shaft current amplitude and distortion rate. The overheated area analysis unit includes a temperature gradient calculation subunit, a first comparative analysis subunit, and a second comparative analysis subunit, wherein: The temperature gradient calculation subunit is used to calculate the axial temperature gradient of the bearing based on the temperature difference between the two ends of the axial direction and the bearing length. The first comparative analysis subunit is used to compare the bearing axial temperature gradient with the temperature gradient threshold, and determine whether there is any abnormality in the bearing's operating state based on the comparison results; The second comparative analysis subunit is used to compare the real-time superheated area with the standard superheated area to obtain the first comparison result and the second comparison result; The data analysis module also includes a corrosion product analysis unit; The corrosion product analysis unit is used to determine the degree of corrosion product accumulation based on the hot spot shape and edge temperature gradient; The fault early warning module includes a stress concentration early warning unit and an electrocorrosion fault early warning unit, wherein, The stress concentration early warning unit is used to trigger a stress concentration early warning prompt based on the degree of accumulation of corrosion products. The electro-corrosion fault early warning unit is used to trigger an electro-corrosion fault early warning based on the bearing wear condition and the duration of temperature, as well as based on the shaft current characteristics. The lubrication failure early warning unit is used to trigger lubrication failure early warning based on shaft current characteristics, the degree of corrosion product accumulation, the duration of temperature, and the rate of temperature change. The latent variable extraction module includes a detection unit, an extraction unit, and an execution unit, wherein: The detection unit is used to detect the temperature of the coupling component based on the degree of corrosion product accumulation and the duration of temperature. The extraction unit is used to extract interference features of the coupling components based on the detection results; The execution unit is used to perform temperature interference processing of the coupling component according to the interference characteristics of the coupling component.

2. The power equipment fault detection system based on a large model according to claim 1, characterized in that, The first comparative analysis subunit determines that the bearing's operating state is abnormal when the bearing's axial temperature gradient is greater than the temperature gradient threshold. When the axial temperature gradient of the bearing is less than or equal to the temperature gradient threshold, the bearing is considered to be in normal operating condition.

3. The power equipment fault detection system based on a large model according to claim 1, characterized in that, The electro-corrosion fault early warning unit includes a first electro-corrosion fault early warning subunit and a second electro-corrosion fault early warning subunit. The first electro-corrosion fault early warning subunit triggers an electro-corrosion fault early warning prompt based on the shaft current characteristics. The second electro-corrosion fault early warning subunit triggers an electro-corrosion fault early warning prompt when it determines that there is no wear and the temperature duration exceeds the duration threshold.

4. The power equipment fault detection system based on a large model according to claim 1, characterized in that, The bearing wear analysis unit performs correlation analysis by calling historical information from the database, including: Collect vibration signals and perform preprocessing; After preprocessing, the sample is determined to be worn or not worn based on kurtosis and sample entropy.

5. The power equipment fault detection system based on a large model according to claim 1, characterized in that, The threshold adjustment module includes a margin calculation unit, a comparison unit, and an adjustment unit, wherein: The margin calculation unit is used to calculate the time difference between the alarm time and the fault time to obtain the warning margin; The comparison unit is used to compare the warning margin with the margin threshold range; The adjustment unit is used to determine whether to adjust the overheated area threshold based on the comparison results.

Images