A service performance sensing system and method for the meshing surface of a large-scale heavy-duty transmission gear

Abstract

The present invention relates to the technical field of gear monitoring, and discloses a service performance sensing system and method for the meshing surface of a large-scale heavy-duty transmission gear. The method includes obtaining historical data of a target gear under different working conditions to establish an evaluation model for the service performance of the gear meshing surface; collecting real-time data of the target gear, and obtaining the current service performance level of the target gear according to the real-time data and the evaluation model; obtaining a lubricating oil pollution state threshold, and dynamically calibrating the threshold according to the current service performance level of the target gear; determining whether the real-time data exceeds the calibrated threshold, and if it exceeds, performing online filtration on the lubricating oil. The present invention can dynamically adjust the lubricating oil pollution state threshold according to the real-time service state of the gear, realize the adaptive detection and treatment of lubricating oil pollution, extend the service life of the gear, and improve the reliability of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of gear monitoring. More specifically, the present invention relates to a sensing system and method for the service performance of the meshing surface of large heavy-duty transmission gears. Background Art

[0002] With the rapid development of industry, gear transmission devices have been widely used in various mechanical equipment. Gear transmission has the advantages of compact structure, large transmission ratio, high efficiency, etc., and plays a key role in the fields of large heavy-duty equipment such as wind turbines, mining machinery, and metallurgical equipment. However, when gears are in long-term service under harsh working conditions, their meshing surfaces are prone to failures such as wear and pitting, resulting in increased equipment vibration, decreased transmission efficiency, and in severe cases, sudden accidents, causing significant economic losses. Therefore, it is crucial to real-time sense the service performance of gears and accurately evaluate their health status to ensure the safe and efficient operation of equipment.

[0003] For example, Chinese Patent with the publication number CN118690577A discloses a method for early warning of bearing wear. By obtaining the actual value of the copper-based abrasive particle content in the oil, and looking up the corresponding relationship between the bearing wear state and the abrasive particle content under different working conditions recorded in the target mapping database, the risk degree of bearing wear is judged and a fault early warning is carried out. This method establishes the correlation between oil abrasive particles and bearing wear, providing a basis for condition assessment. However, the oil sampling detection takes a long time, it is difficult to achieve on-line real-time monitoring, and the materials and working conditions of bearings and gears are quite different, and the characterization ability of oil indicators is limited. Another example is the Chinese Patent with the authorization announcement number CN118110641B, which discloses an intelligent lubrication system for wind turbine gearboxes based on environmental perception. By collecting the environmental and operating data of the gearbox through sensors, a performance degradation model is constructed, and the impurity content of the lubricating oil is monitored. When the pollution exceeds the standard, filtration treatment is carried out, and new oil is automatically supplied when the lubrication is insufficient. This system realizes the on-line monitoring of lubricating oil pollution and liquid level, but lacks the perception of the dynamic response of the meshing surface, and the pertinence and accuracy of service performance assessment need to be improved. Another example is the Chinese Patent with the publication number CN114755007A, which proposes a gear fault diagnosis method. By performing periodic segmentation, interval comparison, and time-domain averaging on the vibration signal, and judging whether the gear is faulty after obtaining the synchronous average data. This method overcomes the influence of rotational speed fluctuation on diagnosis, but does not consider the vibration characteristic differences caused by changes in lubrication state, and has insufficient early diagnosis ability for microscopic failures of the meshing surface.

[0004] In summary, the prior art lacks real-time on-line sensing means for the service state of the gear meshing surface, it is difficult to comprehensively and accurately evaluate the gear health level, the timeliness of lubricating oil pollution monitoring is not high, which affects the performance of lubricating grease, exacerbates tooth surface wear, ignores the correlation between gear service performance changes and lubrication requirements, and the self-adaptability of lubrication state regulation needs to be improved. Summary of the Invention

[0005] To overcome the above defects of the prior art, the present invention provides a service performance sensing system and method for the meshing surface of large heavy-duty transmission gears, conducts collaborative monitoring of multi-physical field parameters for the meshing tooth surface, real-timely characterizes the service state of the gears, and adaptively calibrates the lubricating oil pollution threshold based on this, dynamically optimizes the lubrication state, thereby prolonging the service life of the gears and improving the operation reliability of the system.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A service performance sensing method for the meshing surface of large heavy-duty transmission gears, comprising:

[0008] Obtain the first data of the target gear, and establish a service performance evaluation model for the meshing surface of the gear according to the first data; the first data includes the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the meshing surface of the target gear under different working conditions;

[0009] Collect the real-time conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the meshing surface of the target gear to generate the second data; according to the second data and the service performance evaluation model for the meshing surface of the gear, obtain the current service performance level of the target gear; according to the first data, obtain the lubricating oil pollution state threshold; dynamically calibrate the lubricating oil pollution state threshold according to the current service performance level of the target gear to obtain the calibrated lubricating oil pollution state threshold; determine whether the second data exceeds the calibrated lubricating oil pollution state threshold, if not, continue real-time monitoring; if it exceeds, conduct online filtration of the lubricating oil to remove pollutants.

[0010] Further, the establishment of the service performance evaluation model for the meshing surface of the gear includes:

[0011] Extract the statistical features and morphological features of the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the meshing surface to form a historical feature set; use the historical feature set as the input and the gear service performance level as the output to train a support vector machine (SVM) regression model to obtain the service performance evaluation model for the meshing surface of the gear; the gear service performance level includes four levels: excellent, good, general, and poor.

[0012] Further, the obtaining of the current service performance level of the target gear includes: extracting the statistical features and morphological features of the second data to generate a real-time feature set; according to the real-time feature set and the service performance evaluation model for the meshing surface of the gear, obtain the current service performance level of the target gear.

[0013] Further, the lubricating oil pollution state threshold includes three sub-thresholds, and the three sub-thresholds are conductivity sub-thresholds , sub-threshold of vibration spectrum anomaly and sub-threshold of temperature gradient anomaly ;

[0014] The obtaining of the lubricating oil pollution state threshold includes:

[0015] Calculating the conductivity sub-threshold according to the historical conductivity time series of the meshing surface ;

[0016] Calculating the anomaly degree of the vibration spectrum for each vibration sample point in the historical vibration spectrum time series, obtaining the vibration spectrum anomaly degree sequence, and taking the 90th percentile of the vibration spectrum anomaly degree sequence as the sub-threshold of the vibration spectrum anomaly ;

[0017] Obtaining the sub-threshold of the temperature gradient anomaly according to the historical temperature gradient time series .

[0018] Furthermore, the obtaining of the sub-threshold of the temperature gradient anomaly according to the historical temperature gradient time series includes:

[0019] Calculating the statistical characteristics of the historical temperature gradient time series to form a temperature gradient feature set;

[0020] Using the principal component analysis method to reduce the dimension of the temperature gradient feature set, extracting the principal component scores, and obtaining the low-dimensional temperature gradient features;

[0021] Taking the low-dimensional temperature gradient features as the input, training a support vector data description detection model; defining the temperature gradient anomaly as the distance from the sample points in the historical temperature gradient time series to the hypersphere of the support vector data description detection model, and obtaining the temperature gradient anomaly sequence;

[0022] Performing statistical analysis on the temperature gradient anomaly sequence and taking its 90th percentile as the sub-threshold of the temperature gradient anomaly.

[0023] Furthermore, the calibrated lubricating oil pollution state threshold includes the calibrated conductivity sub-threshold , the calibrated sub-threshold of the vibration spectrum anomaly and the calibrated sub-threshold of the temperature gradient anomaly ;

[0024] The dynamic calibration of the lubricating oil pollution state threshold according to the current service performance level of the target gear includes:

[0025] If the current service performance level of the target gear is excellent, then increase the three sub-thresholds by the proportionality coefficient a to obtain , and ;

[0026] If the current service performance level of the target gear is good, increase the three sub-thresholds by the proportionality coefficient b to obtain , and ;

[0027] If the current service performance level of the target gear is average, keep the three sub-thresholds unchanged;

[0028] If the current service performance level of the target gear is poor, decrease the three sub-thresholds by the proportionality coefficient c to obtain , and ; where a > b > 1.0 > c.

[0029] Furthermore, the determination of whether the second data exceeds the calibrated lubricant contamination state threshold includes:

[0030] Traverse the real-time conductivity time series of the meshing surface, and determine whether each conductivity sample point in the real-time conductivity time series of the meshing surface exceeds the calibrated conductivity sub-threshold , if so, mark the conductivity sample point that exceeds the calibrated conductivity sub-threshold as a conductivity overlimit sample point. If the conductivity overlimit sample points appear continuously and the number exceeds n1, mark M1 = 1, otherwise mark M1 = 0; n1 is the preset threshold for the number of conductivity overlimit samples, and M1 is the conductivity anomaly indication variable;

[0031] Calculate the vibration spectrum anomaly degree of each vibration sample point in the real-time vibration spectrum time series to obtain the real-time vibration spectrum anomaly degree sequence; traverse the real-time vibration spectrum anomaly degree sequence, and determine whether each vibration sample point in the real-time vibration spectrum anomaly degree sequence exceeds the calibrated vibration spectrum anomaly degree sub-threshold , if so, mark the vibration sample point that exceeds the calibrated vibration spectrum anomaly degree sub-threshold as a vibration anomaly sample point. If the vibration anomaly sample points appear continuously and the number exceeds n2, mark M2 = 1, otherwise mark M2 = 0; n2 is the preset threshold for the number of vibration anomaly samples, and M2 is the vibration anomaly indication variable.

[0032] Furthermore, the determination of whether the second data exceeds the calibrated lubricant contamination state threshold also includes:

[0033] Obtain the real-time temperature gradient anomaly degree sequence of the real-time temperature gradient time series, traverse the real-time temperature gradient anomaly degree sequence, and determine whether each temperature sample point in the real-time temperature gradient anomaly degree sequence exceeds the calibrated temperature gradient anomaly degree sub-threshold , if so, mark the temperature sample point that exceeds the calibrated temperature gradient anomaly degree sub-threshold The temperature sample points are marked as temperature anomaly sample points. If the temperature anomaly sample points appear continuously and the number exceeds n3, then mark M3 = 1; otherwise, mark M3 = 0. n3 is the preset temperature anomaly sample quantity threshold, and M3 is the temperature gradient anomaly indication variable.

[0034] If M1 = 1 or M2 = 1 or M3 = 1, it is determined that the second data exceeds the calibrated lubricating oil pollution state threshold; if M1 = 0 and M2 = 0 and M3 = 0, it is determined that the second data does not exceed the calibrated lubricating oil pollution state threshold.

[0035] Further, the obtaining of the real-time temperature gradient anomaly degree sequence of the real-time temperature gradient time series includes:

[0036] Calculate the statistical characteristics of the real-time temperature gradient time series to form a real-time temperature gradient feature set.

[0037] Use the principal component analysis method to reduce the dimension of the real-time temperature gradient feature set to obtain the real-time temperature gradient low-dimensional features.

[0038] According to the real-time temperature gradient low-dimensional features and the support vector data description detection model, obtain the real-time temperature gradient anomaly degree sequence.

[0039] A service performance sensing system for the meshing surface of a large heavy-duty transmission gear, which is used to implement the above-mentioned service performance sensing method for the meshing surface of a large heavy-duty transmission gear. The system includes:

[0040] Evaluation model construction module: used to obtain the first data of the target gear and establish a service performance evaluation model for the gear meshing surface according to the first data. The first data includes the historical meshing surface conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the target gear under different working conditions.

[0041] Service performance evaluation module: used to collect the real-time meshing surface conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the target gear to generate second data; obtain the current service performance level of the target gear according to the second data and the service performance evaluation model for the gear meshing surface.

[0042] Lubricating oil pollution state determination module: used to obtain the lubricating oil pollution state threshold according to the first data, dynamically calibrate the lubricating oil pollution state threshold according to the current service performance level of the target gear to obtain the calibrated lubricating oil pollution state threshold; judge whether the second data exceeds the calibrated lubricating oil pollution state threshold. If it does not exceed, continue real-time monitoring; if it exceeds, perform online filtration on the lubricating oil to remove pollutants.

[0043] An electronic device includes a memory, a central processing unit, and a computer program stored on the memory and executable on the central processing unit. When the central processing unit executes the computer program, it implements the above-mentioned method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear.

[0044] A computer-readable storage medium stores a computer program thereon. When the computer program is executed, it implements the above-mentioned method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] The present invention establishes an evaluation model by obtaining historical data of the gear under different working conditions. Combining with the real-time service data of the gear, it can accurately evaluate the current service performance state of the gear, providing a basis for subsequent lubricating oil pollution detection. Multi-parameter thresholds for the lubricating oil pollution state are introduced, including conductivity, vibration spectrum abnormality degree, temperature gradient abnormality degree, etc., which can comprehensively reflect the pollution degree of the lubricating oil. The lubricating oil pollution threshold is adaptively calibrated according to the real-time service performance level of the gear. When the service performance of the gear decreases, the lubricating oil pollution threshold is correspondingly reduced to improve the detection rate of lubricating oil pollution; when the service performance of the gear is good, the pollution threshold is appropriately increased to avoid overly frequent lubricating oil treatment and save costs. The lubricating oil pollution state is judged in real time. Once it is detected that the lubricating oil pollution exceeds the standard, online filtration treatment is started in time to achieve early detection and early treatment of lubricating oil pollution, delay gear wear, and extend the service life of the gear. Through algorithms such as temperature gradient abnormality degree analysis, abnormal working conditions such as sudden temperature changes can be effectively identified, providing more information for gear service state diagnosis. Description of the Drawings

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is a principle flow chart of a method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear in the present invention;

[0049] Figure 2 It is a method flow chart for obtaining the first data of the target gear in a method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear in the present invention;

[0050] Figure 3Flowchart of the method for obtaining the lubricant contamination state threshold according to the first data in the method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear of the present invention;

[0051] Figure 4 Flowchart of the method for obtaining the sub-threshold of temperature gradient abnormality in the method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear of the present invention;

[0052] Figure 5 Flowchart of the method for dynamically calibrating the lubricant contamination state threshold in the method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear of the present invention;

[0053] Figure 6 Flowchart of the method for obtaining the real-time temperature gradient abnormality sequence of the real-time temperature gradient time series in the method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear of the present invention;

[0054] Figure 7 Functional module diagram of a system for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear in the present invention. Detailed implementation manners

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Embodiment 1

[0057] Please refer to Figure 1 As shown, this embodiment provides a method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear, including:

[0058] Step S1000, obtaining the first data of the target gear, and establishing an evaluation model for the service performance of the gear meshing surface according to the first data; the first data includes the historical meshing surface conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the target gear under different working conditions;

[0059] Further, step S1000 includes:

[0060] Step S1100, obtaining the first data of the target gear, the first data includes the historical meshing surface conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the target gear under different working conditions;

[0061] Further, as Figure 2 shown, step S1100 includes:

[0062] Step S1110: By arranging an array of conductivity sensors on the meshing surface of the target gear, collect the conductivity data of the meshing surface to form a historical meshing surface conductivity time series.

[0063] Step S1120: By arranging acceleration sensors on the bearing seat of the target gear, collect the vibration signals during the operation of the gear, and through spectral analysis, form a historical vibration spectrum time series.

[0064] Step S1130: By arranging an array of infrared temperature sensors at different positions on the meshing surface of the target gear, collect the temperature data of the meshing surface, and calculate the temperature gradient to form a historical temperature gradient time series.

[0065] Specifically, in step S1100, multi-source heterogeneous data of the gear service state is obtained through multi-sensor data fusion. The conductivity of the gear meshing surface reflects the conductive performance of the surface oil film and is closely related to the degree of surface oil pollution. When the oil pollution on the tooth surface intensifies, impurity particles will reduce the insulation of the oil film, resulting in an increase in conductivity. By arranging an array of conductivity sensors, the time evolution law of the conductivity of the meshing surface can be captured. The vibration spectrum of the gear contains rich dynamic characteristic information. When the tooth surface wear intensifies, characteristic frequency components related to the meshing frequency will be excited. By performing spectral analysis on the vibration signal and extracting the time trend of these characteristic frequency components, the tooth surface wear state can be indirectly evaluated. The temperature distribution on the gear meshing surface is closely related to frictional heat. When lubrication fails, it will cause a sharp increase in local temperature and a larger temperature gradient. By collecting the temperature at different positions and calculating the gradient, the change in the lubrication state of the meshing surface can be sensitively perceived. By comprehensively using multi-source information such as conductivity, vibration, and temperature, the internal service state of the gear can be penetrated from three aspects: oil pollution, wear, and lubrication, to achieve comprehensive perception.

[0066] In step S1100, key performance parameters during the gear service process are obtained through multi-sensor data fusion. Conductivity reflects the oil pollution state of the meshing surface and is a direct characterization of oil product deterioration and pollution particles; the vibration spectrum reveals the dynamic characteristics of the gear and is closely related to tooth surface wear; the temperature gradient depicts the lubrication state of the friction pair from a thermodynamic perspective and is a sensitive indicator of lubrication failure. The time variation laws of these three types of data contain the internal mechanism of the evolution of gear service performance and are important inputs for subsequent evaluation models. The collaborative perception of multi-source heterogeneous data expands the limitations of single physical quantity perception and realizes a three-dimensional description of the gear service state, laying a data foundation for comprehensively and accurately evaluating gear performance. At the same time, non-contact detection methods such as conductivity, vibration, and temperature are used to overcome the limitations of traditional manual inspections, realize online monitoring of the gear state, can timely detect signs of performance degradation, and provide a basis for condition-based maintenance decisions.

[0067] Step S1200: Establish an evaluation model for the service performance of the gear meshing surface based on the first data.

[0068] The establishment of the evaluation model for the service performance of the gear meshing surface includes:

[0069] Extract the statistical features and morphological features of the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the meshing surface to form a historical feature set; use the historical feature set as the input and the gear service performance level as the output to train a support vector machine (SVM) regression model to obtain an evaluation model for the service performance of the gear meshing surface; the gear service performance level includes four levels: excellent, good, general, and poor.

[0070] Specifically, in step S1200, a machine learning method is used to explore the correlation law between multi-source time series data and the gear service performance. Perform data preprocessing on the conductivity time series, vibration spectrum time series, and temperature gradient time series, including operations such as outlier removal and data normalization; outliers are usually caused by sensor failures, signal mutations, etc., which will interfere with subsequent feature extraction and model training, and need to be identified and removed using methods such as the 3σ criterion. Data normalization unifies time series data with different dimensions to the same scale, facilitating subsequent feature fusion. Extract the statistical features and morphological features of multi-source time series from both the time domain and frequency domain perspectives. Statistical features include mean, variance, peak-to-peak value, etc., which describe the central tendency and dispersion degree of time series data; morphological features include skewness, kurtosis, etc., which reflect the distribution form of time series data. By comprehensively using these features, the evolution law of time series data can be fully characterized. Finally, the extracted features are input into a support vector machine for regression training. The support vector machine is a pattern recognition method based on statistical learning theory, which realizes the non-linear mapping of the feature space by seeking the optimal classification hyperplane. Using the service performance level scored by experts for typical damage modes as the training target, optimize the parameters of the support vector machine model to establish a non-linear mapping relationship between features and performance levels, forming a service performance evaluation model. This model can quickly estimate the service performance state of the gear based on the input features, providing a quantitative basis for subsequent maintenance decisions. Compared with empirical evaluation, this model makes full use of the degradation law contained in the data, overcomes the limitations of manual experience, and realizes the intelligent perception of service performance.

[0071] In step S1200, a service performance evaluation model for the gear meshing surface is constructed using machine learning methods. With big data-driven as the core concept, this model extracts sensitive features from multi-source heterogeneous data, reveals the degradation mechanism during the gear service process, and breaks through the limitations of traditional physical models. The machine learning model has a powerful non-linear fitting ability and can establish performance degradation patterns under complex working conditions. By learning typical damage cases, the model masters the evaluation knowledge of service performance, can autonomously judge the health state of the gear, and reduces the dependence on manual experience. This evaluation model is a data-driven state perception tool. Through the continuous accumulation and learning of state data, the generalization ability of the model is continuously enhanced, and the evaluation results are more accurate and reliable. Embedding this evaluation model into the online monitoring system can achieve real-time evaluation of the gear service performance and assist maintenance personnel in making scientific decisions. At the same time, the evaluation model is also an effective means of knowledge accumulation. By continuously learning the accumulated data, it can continuously optimize internal parameters, expand the knowledge base of performance degradation, and provide an intelligent tool for the service performance management of the gear throughout its life cycle.

[0072] In step S2000, the real-time conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the meshing surface of the target gear are collected to generate second data; according to the second data and the service performance evaluation model of the gear meshing surface, the current service performance level of the target gear is obtained; according to the first data, the lubricant contamination state threshold is obtained; according to the current service performance level of the target gear, the lubricant contamination state threshold is dynamically calibrated to obtain the calibrated lubricant contamination state threshold; it is judged whether the second data exceeds the calibrated lubricant contamination state threshold. If it does not exceed, continuous real-time monitoring is continued; if it exceeds, step S3000 is executed.

[0073] Furthermore, step S2000 includes:

[0074] In step S2100, the real-time conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the meshing surface of the target gear are collected to generate second data; the statistical features and morphological features of the second data are extracted to generate a real-time feature set;

[0075] Exemplarily, the conductivity data of the gear meshing surface in the most recent 1 hour is collected through a conductivity sensor array, with sampling once per minute, forming a real-time conductivity time series of the meshing surface containing 60 data points;

[0076] The vibration data of the gear in the most recent 1 hour is collected through an acceleration sensor on the bearing housing, with a sampling frequency of 10 kHz. The collected vibration data is subjected to a fast Fourier transform (FFT), and the spectrum data in the range of 0 - 1000 Hz is extracted, with a frequency resolution of 1 Hz, forming a two-dimensional spectrum matrix containing 1000 frequency points and 60 moments as the real-time vibration spectrum time series;

[0077] Collect the temperature data of different positions on the gear meshing surface in the most recent 1 hour through an infrared temperature sensor array, with sampling once per minute. Perform a differential operation on the collected data to obtain the time series of temperature gradients between different positions, forming a real-time temperature gradient time series.

[0078] Extract the statistical features and morphological features of the second data:

[0079] For the real-time conductivity time series of the meshing surface, extract statistical features such as mean, variance, skewness, and kurtosis, and extract morphological features such as trend terms, periodic terms, and random terms.

[0080] For the real-time vibration spectrum time series, extract statistical features such as band energy, spectrum centroid, and spectrum variance, and extract morphological features such as the proportion of sideband, center band, and harmonic band energy.

[0081] For the real-time temperature gradient time series, extract statistical features such as temperature gradient mean and temperature gradient variance, and extract morphological features such as trend mutation points and periodic peak points.

[0082] Summarize the extracted features to form a 25-dimensional real-time feature set.

[0083] Step S2200, according to the real-time feature set and the service performance evaluation model of the gear meshing surface, obtain the current service performance level of the target gear;

[0084] Specifically, in step S2200, the second data obtained by monitoring is input into a pre-trained gear meshing surface service performance evaluation model, which can quickly determine the current service state of the gear. First, the Z-score normalization method is used to dimensionless process each time-series data to improve the fusion ability of data with different physical quantities. The normalization process can unify data with different orders of magnitude to the same level, facilitating subsequent feature extraction and pattern recognition. By extracting the statistical features of the time-series data, the distribution law of the data is comprehensively characterized. The mean and root mean square reflect the central tendency of the data, the variance reflects the degree of dispersion of the data, and the skewness and kurtosis reflect the skewness and kurtosis characteristics of the data. By extracting the morphological features, the evolution law of the time-series data is grasped macroscopically. The time-series decomposition method is used to decompose the time-series data into a trend term, a periodic term, and a random term. The trend term reflects the long-term change trend of the time-series data, the periodic term reflects the periodic fluctuation law in the time-series data, and the random term reflects the random perturbation in the time-series data. The extracted statistical features and morphological features are input into a support vector machine regression model for state recognition, which can automatically determine the service performance level of the gear. The support vector machine maps the features under different damage modes to a high-dimensional space for classification by seeking the maximum margin hyperplane, realizing the intelligent evaluation of service performance. Through the online evaluation of service performance, the health state of the gear can be grasped in a timely manner, realizing the mode transformation from passive maintenance to active prevention, and improving the use reliability and maintenance economy of the gear system.

[0085] Step S2300, according to the first data, obtain the lubricating oil pollution state threshold, and the lubricating oil pollution state threshold includes three sub-thresholds, and the three sub-thresholds are the conductivity sub-threshold , the vibration spectrum abnormality sub-threshold and the temperature gradient abnormality sub-threshold ;

[0086] Further, as Figure 3 shown, step S2300 includes:

[0087] Step S2310, calculate the conductivity sub-threshold according to the historical meshing surface conductivity time series;

[0088]

[0089] Among them:

[0090] : The conductivity sub-threshold represents the critical value of the gear meshing surface conductivity under specific working conditions. This threshold is used to monitor the conductivity change in real time and determine whether a possible fault state has been reached. The higher the threshold, the greater the fluctuation of the surface conductivity and the higher the potential failure risk.

[0091] : The weight coefficient of conductivity volatility, which is used to control the influence of the variance term on the final threshold in the formula. The larger the value of α, the higher the sensitivity to conductivity fluctuations. If the conductivity fluctuations on the gear meshing surface are severe (the variance increases), this term will significantly increase the conductivity sub-threshold , which helps to capture anomalies in a timely manner when the environment is unstable.

[0092] : The weight coefficient of the weighted average of conductivity, which is used to control the influence of the weighted average term on the final threshold in the formula. The weighted average term reflects the average trend of conductivity data as a whole, and β controls the reference value of conductivity under the current working conditions. In some cases, the overall average value or trend of conductivity is more important. When β is larger, the threshold is more dependent on the reference level of conductivity.

[0093] : The weight coefficient of the instantaneous change rate, which is used to control the influence of the maximum difference term on the final threshold in the formula. The maximum difference term is used to capture the sudden changes in conductivity, reflecting the sharp change amplitude of the gear surface conductivity, and γ adjusts the influence of this mutation on the threshold. If γ is larger, it will increase relatively quickly, mainly playing a role in monitoring sudden faults and helping to improve the real-time response ability.

[0094] The value of is obtained by fitting historical data and can usually be set to satisfy , in order to maintain the balance of the overall weight. These coefficients are estimated through machine learning models or optimization algorithms, and the specific values depend on the working conditions and performance requirements of the equipment.

[0095] : The th conductivity data in the historical conductivity time series of the meshing surface.

[0096] : The th conductivity data in the historical conductivity time series of the meshing surface.

[0097] : The average value of conductivity in the historical conductivity time series of the meshing surface, representing the overall level of conductivity during the sampling period.

[0098] : The number of sampling points, that is, the total number of conductivity sample points in the historical conductivity time series of the meshing surface.

[0099] : The The environmental weight factor of conductivity data reflects the impact of the environment on conductivity, such as temperature, humidity, load, etc. This parameter can be calculated through environmental sensor data or combined with existing empirical models to define the influence weight of conductivity under different environmental factors.

[0100] As the conductivity volatility increases, it will increase, indicating that the threshold rises when the fluctuation is large; when the weight factor increases, it means that the current environment has a stronger impact on conductivity, and it will increase accordingly; if the instantaneous conductivity mutates ( significantly), it will also increase to cope with the risk of sudden failure.

[0101] This formula is used to calculate the conductivity sub-threshold under different working conditions, which can achieve efficient monitoring of the conductivity state of the gear meshing surface. This formula comprehensively considers the volatility of conductivity, environmental impact and sudden changes (maximum difference), can keenly capture abnormal states, and is suitable for real-time monitoring in complex environments. The variance term in the formula reflects the conductivity volatility, the weighted average term provides an overall trend judgment, and the maximum difference term is used to identify sudden failures, improving the monitoring accuracy from multiple perspectives. Through and dynamic adjustment, this formula can be flexibly adjusted under different working conditions to adapt to the gear service state evaluation under various working conditions. Through this formula, the fluctuations of gear conductivity under different working conditions can be effectively predicted and responded to, reducing false alarms and enhancing the ability to capture anomalies.

[0102] Step S2320, calculate the vibration spectrum abnormality of each vibration sample point in the historical vibration spectrum time series, obtain the vibration spectrum abnormality sequence, and take the 90th percentile of the vibration spectrum abnormality sequence as the vibration spectrum abnormality sub-threshold;

[0103] Specifically, first, extract the statistical features of the historical vibration spectrum time series, such as mean value, root mean square value, peak factor, waveform factor, etc. These features reflect the energy distribution and waveform characteristics of the vibration signal from both the time domain and the frequency domain. Then, use anomaly detection algorithms such as Mahalanobis distance to calculate the abnormality of each vibration sample. The Mahalanobis distance takes into account the correlation between features and can effectively distinguish normal samples from abnormal samples. The abnormality is positively correlated with the Mahalanobis distance. The larger the abnormality, the greater the degree to which the sample deviates from the normal working condition. By setting an appropriate percentile such as 90%, the abnormality sub-threshold can be adaptively determined, that is, 90% of the historical samples less than this threshold are considered normal. This data-driven threshold determination method makes full use of the statistical laws of historical big data and has better objectivity and robustness.

[0104] Step S2330: Obtain the sub-threshold of temperature gradient abnormality according to the historical temperature gradient time series.

[0105] Further, as Figure 4 shown, step S2330 includes:

[0106] Step S2331: Calculate the statistical features of the historical temperature gradient time series to form a temperature gradient feature set;

[0107] Step S2332: Use the principal component analysis method to reduce the dimension of the temperature gradient feature set, extract the principal component scores, and obtain the low-dimensional temperature gradient features;

[0108] Step S2333: Use the low-dimensional temperature gradient features as input to train a support vector data description detection model; Define the temperature gradient abnormality as the distance from the temperature sample point in the historical temperature gradient time series to the hypersphere of the support vector data description detection model, and obtain a temperature gradient abnormality sequence;

[0109] Step S2334: Conduct statistical analysis on the temperature gradient abnormality sequence, and take its 90% quantile as the sub-threshold of temperature gradient abnormality.

[0110] Specifically, in step S2330, the sub-threshold of temperature gradient abnormality is obtained in a data-driven manner. First, extract features from the historical temperature gradient time series data to obtain statistics reflecting the distribution characteristics of the temperature gradient. For example, the mean reflects the average level of the temperature gradient, the variance reflects the fluctuation range of the temperature gradient, and the skewness and kurtosis reflect the skewness and sharpness of the temperature gradient distribution. Then, use dimension reduction methods such as principal component analysis to remove noise and redundancy while retaining the main feature information of the temperature gradient, and obtain a compact low-dimensional feature representation for facilitating the training of subsequent anomaly detection models.

[0111] The training of the anomaly detection model uses a one-class anomaly detection method based on support vector data description (SVDD). SVDD describes the training samples by finding a hypersphere with the smallest volume, so that the normal samples are enclosed by the hypersphere, while the abnormal samples are outside the hypersphere. When training the SVDD model, only the data in the normal state is required. By minimizing the radius of the hypersphere, the optimal center and radius parameters are obtained. After mapping the temperature sample points to the feature space, calculate the distance from them to the center of the sphere. The larger the distance, the greater the degree of deviation of the temperature sample points from the normal data, that is, the higher the abnormality. Therefore, the temperature gradient abnormality reflects the degree of deviation of the temperature gradient from the normal working condition and is a quantitative index for judging whether the temperature gradient is abnormal.

[0112] All the temperature gradient anomalies are grouped into a temperature gradient anomaly sequence. By statistically analyzing the historical temperature gradient anomaly sequence, a reasonable sub-threshold of the temperature gradient anomaly can be obtained. Using the 90th percentile as the sub-threshold means that the anomalies of 90% of the historical data are lower than this threshold. This threshold can cover most normal operating conditions and has good sensitivity to abnormal situations. During on-line monitoring, calculate the anomaly of the current temperature gradient and compare it with the sub-threshold. If it exceeds the threshold, it can be determined as an abnormal state and alarm processing is required.

[0113] The determination of the anomaly threshold can effectively balance the risks of missed detection and false detection. The selection of a reasonable threshold requires weighing the sensitivity and reliability of detection. Through a large amount of historical data analysis and statistical modeling, the anomaly thresholds for different operating conditions and different equipment can be adaptively determined, improving the accuracy and real-time performance of anomaly detection. The temperature gradient anomaly detection provides a quantitative basis for the temperature anomaly diagnosis and fault warning of the gearbox and is an important means to achieve condition monitoring and fault diagnosis. Combining with the anomaly detection of other state parameters such as vibration and oil quality, a multi-parameter anomaly monitoring system for the gearbox can be constructed to achieve a comprehensive perception and evaluation of the equipment operating state and provide a reliable basis for condition-based maintenance and fault prevention. This method fully utilizes the degradation laws and statistical information contained in the historical temperature gradient data, and obtains an adaptive anomaly quantification index and threshold criterion through data-driven modeling. Compared with the traditional empirical threshold, it can better adapt to the changes in equipment operating conditions and improve the adaptability and robustness of temperature anomaly detection.

[0114] Step S2300 obtains three key threshold indicators of the lubricating oil contamination state through statistical analysis of historical data. The conductivity sub-threshold reflects the critical level of the conductive performance of the oil film on the gear surface and is an important indication of the deterioration and increased contamination degree of the oil product. The determination of this threshold requires comprehensive consideration of the statistical characteristics of the conductivity data, including factors such as data volatility, weighted trend, and instantaneous mutation, and a reasonable threshold is obtained through weighted calculation. The vibration spectrum anomaly sub-threshold characterizes the degree to which the gear vibration signal deviates from the normal operating condition. By extracting the statistical characteristics of the vibration spectrum, such as mean value, variance, peak factor, etc., and performing anomaly detection, the threshold boundary of the vibration spectrum anomaly can be determined. The greater the anomaly, the greater the deviation degree of the vibration signal from the normal operating condition, reflecting the non-linear dynamic response caused by increased tooth surface wear. The temperature gradient anomaly sub-threshold is used to characterize the risk of local temperature rise during meshing caused by lubrication failure. By performing feature extraction and dimensionality reduction processing on the temperature gradient data and using the support vector data description method to learn the compact hypersphere boundary under normal operating conditions, the temperature gradient anomaly can be sensitively captured. The anomaly quantifies the distance of the temperature gradient from the normal boundary and reflects the risk level of lubrication failure.

[0115] Combining the three sub-thresholds forms a complete characterization of the lubricating oil contamination state threshold. This threshold describes the critical level of the lubricating oil contamination state from three perspectives: electrical performance, kinetics, and thermodynamics. When the actually detected conductivity, vibration spectrum abnormality, or temperature gradient abnormality exceeds the corresponding sub-threshold, it can be determined that the lubricating oil contamination state exceeds the standard, and corresponding maintenance strategies need to be taken. This threshold characterization method that fuses multi-source information overcomes the limitations of a single indicator and improves the reliability of lubricating oil contamination monitoring and early warning.

[0116] Step S2300 uses historical data to obtain the three sub-thresholds. The historical data provides rich background information, which can reflect the normal and abnormal states of the equipment under different working conditions. By analyzing this data, accurate thresholds can be established. Historical data can capture the common patterns and abnormal conditions during the long-term operation of the gear. The thresholds calculated using this data can better reflect the actual working conditions. Through the analysis of historical data, the thresholds can be set more accurately, reducing the possibility of false alarms as abnormal under normal circumstances. Historical data helps identify subtle but important changes, ensuring a rapid response to abnormal situations and improving the sensitivity of the monitoring system. As historical data accumulates, the thresholds can be continuously updated and optimized to adapt to changes in the equipment state and environment, improving the robustness of the system.

[0117] Compared with the traditional single-indicator threshold, the lubricating oil contamination state threshold has the following advantages:

[0118] It fuses multi-source heterogeneous data, provides a comprehensive characterization in three aspects of electrical performance, kinetics, and thermodynamics, and the evaluation is more comprehensive and reliable;

[0119] The determination of the sub-thresholds is based on the mining and analysis of historical big data, and the adaptive extraction of the thresholds is realized through technologies such as mathematical statistics and machine learning, overcoming the subjectivity of empirical thresholds.

[0120] In step S2400, according to the current service performance level of the target gear, the lubricating oil contamination state threshold is dynamically calibrated to obtain the calibrated conductivity sub-threshold 、the calibrated vibration spectrum abnormality sub-threshold and the calibrated temperature gradient abnormality sub-threshold , 、 and which constitute the calibrated lubricating oil contamination state threshold;

[0121] Furthermore, as shown in Figure 5 , step S2400 includes:

[0122] In step S2410, if the current service performance level of the target gear is excellent, the three sub-thresholds are increased by the proportionality coefficient a to obtain 、 and ; 1.2<a≤1.5;

[0123] Step S2420: If the current service performance level of the target gear is good, increase the three sub-thresholds according to the proportional coefficient b to obtain , and ; 1.0<b≤1.2;

[0124] Step S2430, if the current service performance level of the target gear is average, the three sub-thresholds are kept unchanged;

[0125] Step S2440: If the current service performance level of the target gear is poor, then reduce the three sub-thresholds according to the proportional coefficient c to obtain , and ;0.6≤c<1.0.

[0126] Specifically, in step S2400, the previously determined lubricant oil contamination state threshold is adaptively calibrated according to the real-time service performance level of the target gear, and the threshold level is dynamically adjusted to match the health status of the gear. When the gear service performance is evaluated as "excellent", it indicates that all performance indicators of the gear are within the ideal range, the tooth surface and oil are in good condition, and the tolerance of the oil contamination threshold can be appropriately increased. By adjusting the proportional coefficient a(1.2 <a≤1.5)放大三个子阈值,扩大阈值的界定范围,在确保充分润滑裕度的同时减少不必要的检修干预,提高设备利用率和运维效率。当齿轮服役性能评估为"良好"时,表明齿轮总体状态正常,但个别指标略有下降,需适度加强监测。通过按比例系数b(1.0<b≤1.2)小幅调高三个子阈值,在保证安全的前提下,为状态变化预留一定的观察窗口期,避免过于敏感的误报警。当齿轮服役性能评估为"一般"时,表明齿轮已经出现了一定程度的性能退化,需按照标准要求严格管控润滑油污染状态,及时实施预防性维护。此时宜维持原有的油污染阈值水平,确保状态监测的裕度。当齿轮服役性能评估为"较差"时,表明齿轮已处于亚健康或故障前期状态,润滑、磨损等问题较为突出,需密切关注其退化趋势。通过按比例系数c(0.6≤c<1.0)缩小三个子阈值,提高对油污染的敏感性,前移预警时间点,争取更多的维护窗口期,避免润滑失效引发的连锁反应。

[0127] The proportionality coefficients a, b, and c for threshold calibration can be empirically set according to factors such as the fault history, lubrication characteristics, and reliability requirements of the gear equipment. Usually, a > b > 1.0 > c follows the principle of "loose entry and strict exit", that is, the threshold is appropriately relaxed for better health conditions, while it is tightened in a timely manner for sub-healthy or high-risk states. The setting of the proportionality coefficients needs to balance safety and economy on the basis of sufficient experimental verification to ensure the rationality of threshold adjustment. For example, for a wind power gearbox, through a large number of fault case analyses, it is found that when the lubricant contamination level is relatively low, the conductivity sub-threshold can be relaxed by 20%, and when the service performance of the gearbox is evaluated as "good" and the load is low, the conductivity sub-threshold can be further relaxed to 40%. However, when the service performance drops to "poor", the conductivity sub-threshold needs to be reduced by 30% to sensitively capture potential faults caused by lubrication deterioration. This adaptive dynamic threshold adjustment strategy can significantly improve the sensitivity and real-time performance of lubricant contamination monitoring, maximize the matching of the equipment's health status, ensure sufficient safety margins while reducing the false alarm rate, and provide a more accurate and reliable decision-making basis for condition-based maintenance.

[0128] Compared with the fixed threshold, the threshold calibration method driven by the service performance level has the following advantages:

[0129] It has strong adaptability and can dynamically adjust the threshold according to the actual health status of the equipment, improving the sensitivity and accuracy of condition monitoring;

[0130] It fully considers the degradation law of the equipment's service performance, takes the performance level as the calibration basis, and the threshold adjustment is more targeted and effective;

[0131] In the slight degradation stage, the threshold is appropriately relaxed to reduce blind maintenance and improve the equipment availability and operation and maintenance efficiency; in the severe degradation stage, the threshold is tightened in a timely manner to advance the early warning time point and gain more time windows for condition-based maintenance decision-making.

[0132] The calibration method is simple and easy to implement. Only by setting 3 proportionality coefficients according to the reliability requirements can the flexible adjustment of the threshold be achieved. The rules are clear and it is easy to apply in engineering.

[0133] Step S2500, continuously determine in real time whether the second data exceeds the calibrated lubricant contamination state threshold. If not, return to step S2100 to continue real-time monitoring; if it exceeds, execute step S3000.

[0134] Furthermore, step S2500 includes:

[0135] Step S2510, traverse the real-time conductivity time series of the meshing surface, and determine whether each conductivity sample point in the real-time conductivity time series of the meshing surface exceeds the calibrated conductivity sub-threshold , if so, the one that exceeds the calibrated conductivity sub-threshold The conductivity sample points are marked as conductivity overlimit sample points. If the conductivity overlimit sample points appear continuously and the number exceeds n1, then mark M1 = 1; otherwise, mark M1 = 0. n1 is the preset threshold for the number of conductivity overlimit samples, and M1 is the conductivity anomaly indication variable.

[0136] Step S2520: Calculate the vibration spectrum anomaly degree of each vibration sample point in the real-time vibration spectrum time series to obtain the real-time vibration spectrum anomaly degree sequence. Traverse the real-time vibration spectrum anomaly degree sequence and determine whether each vibration sample point in the real-time vibration spectrum anomaly degree sequence exceeds the calibrated vibration spectrum anomaly sub-threshold. If so, then the vibration sample points that exceed the calibrated vibration spectrum anomaly sub-threshold are marked as vibration anomaly sample points. If the vibration anomaly sample points appear continuously and the number exceeds n2, then mark M2 = 1; otherwise, mark M2 = 0. n2 is the preset threshold for the number of vibration anomaly samples, and M2 is the vibration anomaly indication variable.

[0137] Step S2530: Obtain the real-time temperature gradient anomaly degree sequence of the real-time temperature gradient time series. Traverse the real-time temperature gradient anomaly degree sequence and determine whether each temperature sample point in the real-time temperature gradient anomaly degree sequence exceeds the calibrated temperature gradient anomaly sub-threshold. If so, then the temperature sample points that exceed the calibrated temperature gradient anomaly sub-threshold are marked as temperature anomaly sample points. If the temperature anomaly sample points appear continuously and the number exceeds n3, then mark M3 = 1; otherwise, mark M3 = 0. n3 is the preset threshold for the number of temperature anomaly samples, and M3 is the temperature gradient anomaly indication variable.

[0138] Furthermore, as Figure 6 shown, step S2530 includes:

[0139] Step S2531: Calculate the statistical features of the real-time temperature gradient time series to form the real-time temperature gradient feature set.

[0140] Step S2532: Use the principal component analysis method to reduce the dimension of the real-time temperature gradient feature set to obtain the real-time temperature gradient low-dimensional features.

[0141] Step S2533: According to the real-time temperature gradient low-dimensional features and the support vector data description detection model, obtain the real-time temperature gradient anomaly degree sequence.

[0142] Specifically, in step S2531, statistical feature extraction is performed on the time series of the temperature gradient collected in real time. The temperature gradient reflects the non-uniformity of the heat distribution on the gear meshing surface and is closely related to the lubrication state. Commonly used statistical features include mean, variance, peak-to-peak value, root mean square, etc. The mean reflects the overall level of the temperature gradient, the variance reflects the fluctuation degree of the temperature gradient, the peak-to-peak value reflects the maximum change range of the temperature gradient, and the root mean square comprehensively reflects the energy size of the temperature gradient. These features characterize the distribution law of the temperature gradient time series from different aspects and form a real-time temperature gradient feature set. However, the dimension of the feature set is relatively high, containing a large amount of redundant and noisy information, which is not conducive to subsequent anomaly detection. In step S2532, the principal component analysis (PCA) method is used to reduce the dimension of the real-time temperature gradient feature set. PCA is a commonly used linear dimensionality reduction algorithm that maps the original high-dimensional features to a low-dimensional space through orthogonal transformation and extracts the main structural information of the data. PCA selects several orthogonal directions with the largest variance as the principal components according to the variance size of the features, discards the components with smaller variance, and realizes data compression and noise reduction. By reducing the dimension with PCA, not only the computational cost of the feature set is reduced, but also the key information contained in the temperature gradient is highlighted, laying a foundation for subsequent anomaly detection. The data after dimensionality reduction is called the low-dimensional feature of the real-time temperature gradient.

[0143] In step S2533, the anomaly degree of the temperature gradient is calculated using the low-dimensional feature of the real-time temperature gradient and the pre-trained support vector data description (SVDD) model. SVDD is a one-class anomaly detection algorithm that judges data points deviating from the hypersphere as anomalies by learning the compact hypersphere of the data in the normal state. The anomaly degree can be measured by the distance from the data point to the hypersphere. The larger the distance, the greater the degree of deviation of the data point from the normal state. In order to adapt to the dynamic changes of the temperature gradient, a sliding time window method is used for anomaly detection. The anomaly degree of the temperature gradient within a certain time range before and after the current moment is calculated each time, forming a real-time temperature gradient anomaly degree sequence. This sequence dynamically reflects the change trend of the temperature gradient and provides a quantitative index for the real-time monitoring of the lubrication state.

[0144] Through steps S2531 - S2533, the original temperature gradient time series undergoes a series of processes such as feature extraction, principal component analysis, and anomaly detection, and is transformed into a real - time temperature gradient anomaly degree sequence. This anomaly degree sequence eliminates redundant information and random noise in the temperature gradient, extracts core features closely related to the lubrication state, and quantifies the degree of deviation of the temperature gradient from the normal state by comparing with the SVDD model. The real - time generation of the anomaly degree sequence provides a new idea for online lubrication state assessment. By setting an appropriate anomaly degree threshold, potential risks of lubrication failure can be detected in a timely manner, providing a reliable basis for condition - based maintenance decisions. Compared with the monitoring of a single physical quantity, the temperature gradient anomaly degree makes full use of the information of the temperature spatial distribution, and has higher sensitivity and diagnostic accuracy for the lubrication state. Integrating this anomaly degree index into the gear health monitoring system can realize real - time assessment and early warning of the lubrication state, reduce the probability of accidents, and improve the reliability and safety of the gear transmission system.

[0145] In step S2540, if M1 = 1 or M2 = 1 or M3 = 1, it is determined that the second data exceeds the calibrated lubricating oil contamination state threshold; if M1 = 0 and M2 = 0 and M3 = 0, it is determined that the second data does not exceed the calibrated lubricating oil contamination state threshold.

[0146] Specifically, step S2500 comprehensively uses multi - source heterogeneous data such as conductivity, vibration spectrum, and temperature gradient, and determines whether the gear meshing surface is in the lubricating oil contamination state by comparing with the corresponding calibrated thresholds. In step S2510, the real - time conductivity time series of the meshing surface is traversed, and each conductivity sample point is judged one by one whether it exceeds the calibrated conductivity sub - threshold. Conductivity is an important indicator reflecting the degree of lubricating oil contamination. When the impurity particles in the lubricating oil increase, it will cause the conductivity of the oil film to increase and the conductivity to rise. Therefore, the conductivity exceeding the threshold is an important sign of the aggravation of lubricating oil contamination. To improve the reliability of detection and avoid the interference of single abnormal points, this step sets the criterion for continuous abnormal points, that is, only when the over - limit sample points appear continuously and the number exceeds n1, it is considered that conductivity anomaly has occurred, and M1 is marked as 1. Here, n1 represents the number threshold of consecutive occurrences of conductivity over - limit sample points in the real - time conductivity time series of the meshing surface. That is, when the number of consecutive occurrences of conductivity over - limit sample points exceeds n1, it is considered that conductivity anomaly has occurred. n1 is an empirical value, usually taken as 10% - 20% of the total length of the time series. For example, for a time series with a length of 1000, n1 can be taken between 100 - 200. The setting of n1 needs to balance sensitivity and reliability. Too small will lead to an increase in false alarms, and too large may lead to missed detections. M1 is the conductivity anomaly indication variable. When the over - limit sample points of conductivity appear continuously and the number exceeds n1, M1 = 1, indicating conductivity anomaly; otherwise, M1 = 0, indicating normal conductivity.

[0147] The idea of step S2520 is similar to that of step S2510, except that the criterion object is changed to the vibration spectrum abnormality. The vibration spectrum reflects the dynamic characteristics of the gear. When the tooth surface wear intensifies, characteristic frequency components will appear near the meshing frequency, manifested as local abnormalities in the spectrum. By extracting the abnormality characteristics of the real-time spectrum and comparing them with the calibration threshold, the tooth surface wear state can be judged. Similar to the conductivity, the vibration abnormality also needs to occur continuously for a certain number (>n2) to be confirmed, so as to improve the reliability. n2 represents the number threshold of consecutive occurrences of vibration abnormality sample points in the real-time vibration spectrum abnormality sequence. That is, when the number of consecutive occurrences of vibration abnormality sample points exceeds n2, the vibration is considered abnormal. M2 is the vibration abnormality indication variable. When the vibration abnormality sample points occur continuously and the number exceeds n2, M2 = 1, indicating vibration abnormality; otherwise, M2 = 0, indicating normal vibration.

[0148] Step S2530 is for the judgment of the temperature gradient abnormality. The temperature gradient reflects the uniformity of the heat distribution on the tooth surface and is an important indicator of the lubrication state. When the lubrication fails, it will cause local frictional heat generation and a sharp increase in the temperature gradient. Therefore, when the temperature gradient abnormality exceeds the threshold, it means that the lubricating oil may have failed. Similarly, the temperature abnormality needs to occur continuously for a sufficient number of times (>n3) to be confirmed. n3 represents the number threshold of consecutive occurrences of temperature abnormality sample points in the real-time temperature gradient abnormality sequence. That is, when the number of consecutive occurrences of temperature abnormality sample points exceeds n3, the temperature gradient is considered abnormal. M3 is the temperature gradient abnormality indication variable. When the temperature abnormality sample points occur continuously and the number exceeds n3, M3 = 1, indicating temperature gradient abnormality; otherwise, M3 = 0, indicating normal temperature gradient.

[0149] Step S2540 is to synthesize the above three judgment results to obtain the final lubricating oil contamination state. Since the conductivity, vibration, and temperature reflect the health level of the lubricating oil from different aspects, as long as any one item is judged to be abnormal, it is considered that the lubricating oil has been contaminated and maintenance measures need to be taken. Only when all three criteria are normal can the health of the lubricating oil state be confirmed. This combined criterion helps to improve the reliability of detection and reduce false negatives and false positives.

[0150] In summary, step S2500 makes full use of multi-source heterogeneous data to comprehensively reflect the lubricating oil contamination state, avoiding the limitations of single information; adopts a relative threshold criterion to overcome the influence of working condition fluctuations and can adapt to different service conditions; integrates the detection results of multiple subsystems to improve the accuracy and reliability of diagnosis; can perform online real-time monitoring to timely detect signs of lubricating oil contamination and facilitate condition-based maintenance.

[0151] Exemplarily, in step S2510, traverse the real-time meshing surface conductivity time series {0.08, 0.09, 0.12, 0.15, 0.11, 0.13, 0.12, 0.14, 0.18, 0.19, 0.21, 0.14, 0.16, 0.15, 0.16}, and assume that the calibrated conductivity sub-threshold is 0.16. Determine whether each conductivity sample point in the real-time meshing surface conductivity time series exceeds the calibrated conductivity sub-threshold. For the 9th, 10th, and 11th points that exceed the threshold of 0.16, mark them as conductivity over-threshold sample points. If 3 conductivity over-threshold sample points appear continuously and exceed the given quantity n1 (assuming n1 = 2), then mark M1 = 1.

[0152] In step S2520, calculate the vibration spectrum abnormality degree of each vibration sample point in the real-time vibration spectrum time series, and obtain the real-time vibration spectrum abnormality degree sequence {0.12, 0.15, 0.22, 0.18, 0.14, 0.08, 0.13, 0.09, 0.17, 0.19, 0.20, 0.26, 0.29, 0.27, 0.28}, and assume that the calibrated vibration spectrum abnormality sub-threshold is 0.25. Traverse the real-time vibration spectrum abnormality degree sequence, and determine whether each vibration sample point in it exceeds the calibrated vibration spectrum abnormality sub-threshold. For the 12th, 13th, 14th, and 15th points that exceed the threshold of 0.25, mark them as vibration abnormality sample points. If 4 vibration abnormality sample points appear continuously and exceed the given quantity n2 (assuming n2 = 3), then mark M2 = 1.

[0153] In step S2530, obtain the real-time temperature gradient abnormality degree sequence {0.02, 0.05, 0.03, 0.06, 0.09, 0.13, 0.12, 0.15, 0.18, 0.04, 0.10, 0.09, 0.07, 0.06, 0.04} of the real-time temperature gradient time series, and assume that the calibrated temperature gradient abnormality sub-threshold is 0.10. Traverse the real-time temperature gradient abnormality degree sequence, and determine whether each temperature sample point in it exceeds the calibrated temperature gradient abnormality sub-threshold. For the 6th, 7th, 8th, and 9th points that exceed the threshold of 0.10, mark them as temperature abnormality sample points. If 4 temperature abnormality sample points appear continuously but do not exceed the given quantity n3 (assuming n3 = 5), then mark M3 = 0.

[0154] In step S2540, since M1 = 1, M2 = 1, and M3 = 0, satisfying "M1 = 1 or M2 = 1 or M3 = 1", it is determined that the second data exceeds the calibrated lubricant contamination state threshold, and step S3000 needs to be executed.

[0155] This example simulates the possible situations of conductivity, vibration spectrum abnormality degree, and temperature gradient abnormality degree under actual working conditions. By judging the number of abnormal points in each time-series data, marking M1, M2, and M3, and finally deciding whether it exceeds the calibrated lubricating oil pollution state threshold according to the marked values. This method of comprehensive judgment of multiple parameters can improve the reliability of oil condition monitoring.

[0156] Generally speaking, step S2000 integrates multi-source sensing information such as conductivity, vibration, and temperature, and constructs a multi-dimensional monitoring system for the service performance of the gear meshing surface. By collecting real-time data of the target gear, calculating and integrating the abnormality degree indicators corresponding to each physical quantity, the dynamic changes of the gear service state can be accurately grasped, and the early signs of performance degradation can be detected in time. This data-driven online monitoring mode breaks through the traditional single fault diagnosis method based on thresholds and realizes the active perception of the gear service state. The monitoring results are output in the form of service performance levels, intuitively reflecting the health level of the gear, which is convenient for engineers to quickly understand and make decisions. At the same time, by optimizing the evaluation model parameters, the sensitivity and accuracy of monitoring can be further improved, and the risk of false alarms and missed alarms can be reduced. Applying this solution to the gear drive systems of wind power, rail transit and other equipment can significantly improve the safety and reliability of the system, maximize the service life of the equipment, and reduce the operation and maintenance costs. In addition, this solution also lays a data foundation for gear fault warning and remaining life prediction, and has important value for realizing the reliability design and intelligent operation and maintenance of gears.

[0157] In step S3000, the lubricating oil is filtered online to remove contaminants.

[0158] Specifically, in step S3000, when the monitoring data exceeds the calibrated lubricating oil pollution state threshold, it is necessary to filter the lubricating oil online to remove contaminants and restore the cleanliness and lubrication performance of the lubricating oil. Online filtration refers to a method of directly purifying the lubricating oil without stopping the machine during the operation of the equipment. Its main purpose is to timely remove harmful substances such as pollution particles, moisture, and oxidation products in the lubricating oil, maintain the performance indicators of the lubricating oil, extend the service life of the lubricating oil, and ensure the normal operation of the equipment.

[0159] The online filtration system usually consists of multi-stage precision filters, heaters, oil pumps, controllers, etc. The precision filter adopts a multi-layer structure, and the filtration accuracy gradually increases from the inside to the outside, which can effectively intercept pollution particles of different sizes. For example, the commonly used β (5)1000The filter element with high precision means that for pollution particles with a size greater than or equal to 5μm, its interception rate can reach over 99.9%. The selection of the filter material is also crucial. Glass fiber filter material has a good filtering effect on solid particles, while coalescing filter material can effectively remove moisture. The heater can heat the lubricating oil to an appropriate temperature, reduce the viscosity of the oil, and accelerate the flow of the oil and the separation of impurities. The oil pump provides stable circulating power to ensure the filtering efficiency. The controller can automatically adjust the working state of the filtering system according to parameters such as the pollution degree, temperature, and flow rate of the oil.

[0160] The working principle of on-line filtration is that through the way of bypass circulation, a part of the lubricating oil is introduced into the filtration system. After multi-stage precision filtration, it then returns to the main oil circuit to achieve continuous purification of the oil. A pollution degree sensor is installed at the inlet of the filter to monitor the pollution state of the lubricating oil in real time. When the pollution degree exceeds the standard, the system automatically switches to the filtration mode. At the same time, the clogging state of the filter is also monitored in real time. When the pressure difference exceeds the set value, the controller issues a warning to remind to replace the filter element in time. The whole filtration process runs automatically without affecting the normal operation of the main equipment, and can significantly improve the operation efficiency and reliability of the equipment.

[0161] On-line filtration technology has significant technical advantages and economic benefits. Compared with off-line filtration, on-line filtration does not interrupt the operation of the equipment, can continuously maintain the cleanliness of the lubricating oil, and reduce production losses caused by shutdown. At the same time, the on-line filtration system is equipped with precision filters with high filtration accuracy and high purification efficiency. A large amount of lubricating oil can be filtered to the ideal cleanliness within 1 hour. Compared with replacing new oil, on-line filtration greatly extends the service life of the lubricating oil, saves the storage and disposal costs, and has significant economic benefits. In addition, on-line filtration has precise control and can be adjusted in real time according to the pollution state of the lubricating oil, avoiding waste of filter materials and energy consumption caused by over-filtration.

[0162] In summary, the on-line filtration technology in step S3000 is an efficient, economic, and environmentally friendly lubricating oil purification method. Through multi-stage precision filtration and real-time monitoring, the quality of the lubricating oil can be continuously improved, and the service performance and reliability of the gear transmission device can be enhanced. The reasonable design and optimized control of the on-line filtration device are the key links to realize intelligent condition monitoring and maintenance, and are of great significance for improving the equipment level and economic benefits.

[0163] The present invention can be applied to large-scale heavy-duty transmission gear systems such as wind power gearboxes and industrial transmission devices, and has broad application prospects. Compared with the traditional method of regular detection and replacement of lubricating oil, the present invention can significantly improve the timeliness and pertinence of detection, reduce the detection and maintenance costs, improve the equipment reliability, and is of great significance for promoting the safe and economic operation of large-scale transmission devices.

[0164] Embodiment 2

[0165] Based on Embodiment 1, this embodiment provides a sensing system for the service performance of the meshing surface of a large-scale heavy-duty transmission gear, as follows Figure 7 shown, including:

[0166] Evaluation model construction module: used to obtain the first data of the target gear, and establish an evaluation model for the service performance of the gear meshing surface according to the first data; the first data includes the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the meshing surface of the target gear under different working conditions;

[0167] Service performance evaluation module: used to collect the real-time conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the meshing surface of the target gear, and generate the second data; according to the second data and the evaluation model for the service performance of the gear meshing surface, obtain the current service performance level of the target gear;

[0168] Lubricating oil pollution state determination module: used to obtain the lubricating oil pollution state threshold according to the first data, dynamically calibrate the lubricating oil pollution state threshold according to the current service performance level of the target gear, and obtain the calibrated lubricating oil pollution state threshold; determine whether the second data exceeds the calibrated lubricating oil pollution state threshold. If it does not exceed, continue to monitor in real time; if it exceeds, perform online filtration of the lubricating oil to remove pollutants.

[0169] In the evaluation model construction module, the obtaining of the first data of the target gear includes:

[0170] Step S1110, by arranging a conductivity sensor array on the meshing surface of the target gear, collecting the conductivity data of the meshing surface, and forming a historical conductivity time series of the meshing surface;

[0171] Step S1120, by arranging an acceleration sensor on the bearing seat of the target gear, collecting the vibration signal during the operation of the gear, and forming a historical vibration spectrum time series through spectrum analysis;

[0172] Step S1130, by arranging an infrared temperature sensor array at different positions on the meshing surface of the target gear, collecting the temperature data of the meshing surface, and calculating the temperature gradient to form a historical temperature gradient time series.

[0173] The establishment of the evaluation model for the service performance of the gear meshing surface includes:

[0174] Extract the statistical features and morphological features of the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the meshing surface to form a historical feature set; use the historical feature set as the input and the gear service performance level as the output to train a support vector machine (SVM) regression model to obtain the evaluation model for the service performance of the gear meshing surface; the gear service performance level includes four levels: excellent, good, average, and poor.

[0175] In the lubricating oil pollution state determination module, obtaining the lubricating oil pollution state threshold according to the first data includes:

[0176] Step S2310: Calculate the conductivity sub-threshold according to the historical conductivity time series of the meshing surface;

[0177] Step S2320: Calculate the vibration spectrum abnormality degree of each vibration sample point in the historical vibration spectrum time series, obtain the vibration spectrum abnormality degree sequence, and take the 90th percentile of the vibration spectrum abnormality degree sequence as the vibration spectrum abnormality sub-threshold;

[0178] Step S2330: Obtain the temperature gradient abnormality sub-threshold according to the historical temperature gradient time series.

[0179] The said step S2330 includes:

[0180] Step S2331: Calculate the statistical features of the historical temperature gradient time series to form a temperature gradient feature set;

[0181] Step S2332: Use the principal component analysis method to reduce the dimension of the temperature gradient feature set, extract the principal component scores, and obtain the temperature gradient low-dimensional features;

[0182] Step S2333: Use the temperature gradient low-dimensional features as the input to train the support vector data description detection model; define the temperature gradient abnormality as the distance from the temperature sample point in the historical temperature gradient time series to the hypersphere of the support vector data description detection model, and obtain the temperature gradient abnormality sequence;

[0183] Step S2334: Conduct statistical analysis on the temperature gradient abnormality sequence, and take its 90th percentile as the temperature gradient abnormality sub-threshold.

[0184] In the lubricating oil pollution state determination module, the calibrated lubricating oil pollution state threshold includes the calibrated conductivity sub-threshold , the calibrated vibration spectrum abnormality sub-threshold and the calibrated temperature gradient abnormality sub-threshold ;

[0185] The dynamic calibration of the lubricating oil pollution state threshold includes:

[0186] Step S2410: If the current service performance level of the target gear is excellent, increase the three sub-thresholds by the proportionality coefficient a to obtain , and ; 1.2 < a ≤ 1.5;

[0187] Step S2420: If the current service performance level of the target gear is good, increase the three sub-thresholds by the proportionality coefficient b to obtain , and ; 1.0 < b ≤ 1.2;

[0188] Step S2430, if the current service performance level of the target gear is average, then keep the three sub-thresholds unchanged;

[0189] Step S2440, if the current service performance level of the target gear is poor, then reduce the three sub-thresholds by the proportionality coefficient c to obtain , and ; 0.6 ≤ c < 1.0.

[0190] In the lubricant contamination state determination module, the real-time judgment of whether the second data exceeds the calibrated lubricant contamination state threshold includes:

[0191] Step S2510, traverse the real-time conductivity time series of the meshing surface, and judge whether each conductivity sample point in the real-time conductivity time series of the meshing surface exceeds the calibrated conductivity sub-threshold , if so, mark the conductivity sample point that exceeds the calibrated conductivity sub-threshold as a conductivity overlimit sample point. If the conductivity overlimit sample points appear continuously and the number exceeds n1, then mark M1 = 1, otherwise mark M1 = 0; n1 is the preset conductivity overlimit sample quantity threshold, and M1 is the conductivity anomaly indication variable;

[0192] Step S2520, calculate the vibration spectrum anomaly degree of each vibration sample point in the real-time vibration spectrum time series to obtain the real-time vibration spectrum anomaly degree sequence; traverse the real-time vibration spectrum anomaly degree sequence, and judge whether each vibration sample point in the real-time vibration spectrum anomaly degree sequence exceeds the calibrated vibration spectrum anomaly degree sub-threshold , if so, mark the vibration sample point that exceeds the calibrated vibration spectrum anomaly degree sub-threshold as a vibration anomaly sample point. If the vibration anomaly sample points appear continuously and the number exceeds n2, then mark M2 = 1, otherwise mark M2 = 0; n2 is the preset vibration anomaly sample quantity threshold, and M2 is the vibration anomaly indication variable;

[0193] Step S2530, obtain the real-time temperature gradient anomaly degree sequence of the real-time temperature gradient time series, traverse the real-time temperature gradient anomaly degree sequence, and judge whether each temperature sample point in the real-time temperature gradient anomaly degree sequence exceeds the calibrated temperature gradient anomaly degree sub-threshold , if so, mark the temperature sample point that exceeds the calibrated temperature gradient anomaly degree sub-threshold The temperature sample points are marked as temperature anomaly sample points. If the temperature anomaly sample points appear continuously and the quantity exceeds n3, then mark M3 = 1; if not, then mark M3 = 0. n3 is a preset temperature anomaly sample quantity threshold, and M3 is a temperature gradient anomaly indication variable.

[0194] Step S2540: If M1 = 1 or M2 = 1 or M3 = 1, then determine that the second data exceeds the calibrated lubricating oil pollution state threshold; if M1 = 0 and M2 = 0 and M3 = 0, then determine that the second data does not exceed the calibrated lubricating oil pollution state threshold.

[0195] The said step S2530 includes:

[0196] Step S2531: Calculate the statistical features of the real-time temperature gradient time series to form a real-time temperature gradient feature set.

[0197] Step S2532: Use the principal component analysis method to reduce the dimension of the real-time temperature gradient feature set to obtain the real-time temperature gradient low-dimensional features.

[0198] Step S2533: According to the real-time temperature gradient low-dimensional features and the support vector data description detection model, obtain the real-time temperature gradient anomaly degree sequence.

[0199] Embodiment 3

[0200] This embodiment discloses an electronic device, which may include one or more processors and one or more memories. Among them, computer-readable code is stored in the memory, and when the computer-readable code is run by one or more processors, it can execute a method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear as described above.

[0201] The method or system according to the embodiments of the present application can also be implemented by means of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, input / output components, a hard disk, etc. The storage device in the electronic device, such as the ROM or the hard disk, can store a method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear provided by the present application. A method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear may, for example, include: obtaining first data of a target gear, and establishing an evaluation model for the service performance of the gear meshing surface according to the first data; the first data includes the historical conductivity time series of the meshing surface of the target gear under different working conditions, the historical vibration spectrum time series, and the historical temperature gradient time series; collecting the real-time conductivity time series, the real-time vibration spectrum time series, and the real-time temperature gradient time series of the meshing surface of the target gear to generate second data; obtaining the current service performance level of the target gear according to the second data and the evaluation model for the service performance of the gear meshing surface; obtaining a lubricating oil contamination state threshold according to the first data; dynamically calibrating the lubricating oil contamination state threshold according to the current service performance level of the target gear to obtain a calibrated lubricating oil contamination state threshold; determining whether the second data exceeds the calibrated lubricating oil contamination state threshold, if not, continuing real-time monitoring; if so, performing online filtration on the lubricating oil to remove contaminants.

[0202] Further, the electronic device may further include a user interface. Of course, the architecture disclosed in the present invention is only exemplary. When implementing different devices, one or more components of the electronic device disclosed in the present invention may be omitted according to actual needs.

[0203] Embodiment 4

[0204] This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are run by a processor, a method for sensing the service performance of the meshing surface of a large heavy-duty transmission gear according to the embodiments of the present application can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may, for example, include random access memory (RAM) and cache memory, etc. Non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0205] In addition, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be run by a processor to execute instructions corresponding to the method steps provided by the present application. For example: obtaining first data of a target gear, and establishing a service performance evaluation model for the gear meshing surface according to the first data; the first data includes the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the target gear under different working conditions; collecting the real-time conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the target gear meshing surface to generate second data; obtaining the current service performance level of the target gear according to the second data and the service performance evaluation model for the gear meshing surface; obtaining a lubricating oil pollution state threshold according to the first data; dynamically calibrating the lubricating oil pollution state threshold according to the current service performance level of the target gear to obtain a calibrated lubricating oil pollution state threshold; determining whether the second data exceeds the calibrated lubricating oil pollution state threshold, if not, continuing real-time monitoring; if so, performing online filtration on the lubricating oil to remove pollutants. When this computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0206] The methods, systems, and devices of the present application can be implemented in many ways. For example, the methods, systems, and devices of the present application can be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present application are not limited to the specific order described above, unless otherwise specifically stated. In addition, in some embodiments, the present application can also be implemented as a program recorded in a recording medium, and these programs include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers a recording medium storing a program for executing the method according to the present application.

[0207] In addition, parts of the above technical solutions provided in the embodiments of the present application that are consistent with the corresponding technical solutions in the prior art in terms of implementation principles are not described in detail to avoid excessive elaboration.

[0208] As described above in the specific embodiments, the purpose, technical solutions, and beneficial effects of the present invention are further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear, characterized in that The method includes: Obtaining first data of a target gear, and establishing a service performance evaluation model for the gear meshing surface according to the first data; the first data includes the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the target gear under different working conditions; Collecting the real-time conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the target gear to generate second data; obtaining the current service performance level of the target gear according to the second data and the service performance evaluation model for the gear meshing surface; obtaining a lubricating oil pollution state threshold according to the first data; dynamically calibrating the lubricating oil pollution state threshold according to the current service performance level of the target gear to obtain a calibrated lubricating oil pollution state threshold; determining whether the second data exceeds the calibrated lubricating oil pollution state threshold, if not, continue to monitor in real time; if it exceeds, perform online filtration on the lubricating oil to remove pollutants; The calibrated lubricating oil contamination state threshold includes a calibrated conductivity sub-threshold , a calibrated vibration spectrum anomaly sub-threshold and a calibrated temperature gradient anomaly sub-threshold ; The method for determining whether the second data exceeds the calibrated lubricating oil pollution state threshold includes: Identifying conductivity overlimit sample points in the real-time conductivity time series of the meshing surface. If the conductivity overlimit sample points appear continuously and the number exceeds n1, mark M1 = 1, otherwise mark M1 = 0; n1 is a preset conductivity overlimit sample quantity threshold, and M1 is a conductivity anomaly indication variable; identifying vibration anomaly sample points in the real-time vibration spectrum time series. If the vibration anomaly sample points appear continuously and the number exceeds n2, mark M2 = 1, otherwise mark M2 = 0; n2 is a preset vibration anomaly sample quantity threshold, and M2 is a vibration anomaly indication variable; identifying temperature anomaly sample points in the real-time temperature gradient time series. If the temperature anomaly sample points appear continuously and the number exceeds n3, mark M3 = 1, otherwise mark M3 = 0; n3 is a preset temperature anomaly sample quantity threshold, and M3 is a temperature gradient anomaly indication variable; if M1 = 1 or M2 = 1 or M3 = 1, it is determined that the second data exceeds the calibrated lubricating oil pollution state threshold; if M1 = 0 and M2 = 0 and M3 = 0, it is determined that the second data does not exceed the calibrated lubricating oil pollution state threshold.

2. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 1, characterized in that, The establishment of the service performance evaluation model for the gear meshing surface includes: Extracting the statistical features and morphological features of the historical conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the meshing surface to form a historical feature set; using the historical feature set as the input and the gear service performance level as the output to train a support vector machine (SVM) regression model to obtain the service performance evaluation model for the gear meshing surface; the gear service performance level includes four levels: excellent, good, general, and poor.

3. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 1, characterized in that, Obtaining the current service performance level of the target gear includes: extracting the statistical features and morphological features of the second data to generate a real-time feature set; obtaining the current service performance level of the target gear according to the real-time feature set and the service performance evaluation model for the gear meshing surface.

4. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 2, characterized in that, The lubricating oil pollution state threshold includes three sub-thresholds, and the three sub-thresholds are the conductivity sub-threshold , the vibration spectrum abnormality sub-threshold , and the temperature gradient abnormality sub-threshold ; Obtaining the lubricating oil pollution state threshold includes: Calculate the conductivity sub-threshold based on the historical time series of the conductivity of the meshing surface ; Calculate the vibration spectrum abnormality degree of each vibration sample point in the historical vibration spectrum time series, obtain the vibration spectrum abnormality degree sequence, and take the 90% quantile of the vibration spectrum abnormality degree sequence as the vibration spectrum abnormality sub-threshold ; Obtain the sub-threshold of temperature gradient anomaly according to the historical temperature gradient time series .

5. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 4, characterized in that, Obtaining a temperature gradient anomaly sub-threshold according to the historical temperature gradient time series includes: Calculating the statistical features of the historical temperature gradient time series to form a temperature gradient feature set; The principal component analysis method is used to reduce the dimension of the temperature gradient feature set, extract the principal component scores, and obtain the low-dimensional temperature gradient features; Using the low-dimensional temperature gradient features as the input, a support vector data description detection model is trained; the temperature gradient anomaly degree is defined as the distance from the sample points in the historical temperature gradient time series to the hypersphere of the support vector data description detection model, and a temperature gradient anomaly degree sequence is obtained; Statistical analysis is performed on the temperature gradient anomaly degree sequence, and its 90% quantile is taken as the sub-threshold of the temperature gradient anomaly degree.

6. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 5, characterized in that, The dynamic calibration of the lubricating oil pollution state threshold according to the current service performance level of the target gear includes: If the current service performance level of the target gear is excellent, increase the three sub-thresholds by the proportionality coefficient a to obtain , and ; If the current service performance level of the target gear is good, increase the three sub-thresholds by the proportionality coefficient b to obtain , and ; If the current service performance level of the target gear is average, the three sub-thresholds remain unchanged; If the current service performance level of the target gear is poor, then reduce the three sub-thresholds by the proportionality coefficient c to obtain , and ; where a > b > 1.0 > c.

7. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 6, characterized in that The method for identifying the conductivity over-limit sample points in the real-time meshing surface conductivity time series is as follows: traverse the real-time meshing surface conductivity time series, and determine whether each conductivity sample point in the real-time meshing surface conductivity time series exceeds the calibrated conductivity sub-threshold , if so, mark the conductivity sample point that exceeds the calibrated conductivity sub-threshold as the conductivity over-limit sample point; The method for identifying temperature anomaly sample points in the real-time temperature gradient time series is: calculating the vibration spectrum anomaly degree of each vibration sample point in the real-time vibration spectrum time series to obtain a real-time vibration spectrum anomaly degree sequence; Traverse the real-time vibration spectrum abnormality degree sequence, and determine whether each vibration sample point in the real-time vibration spectrum abnormality degree sequence exceeds the calibrated vibration spectrum abnormality sub-threshold , if so, then the vibration sample point that exceeds the calibrated vibration spectrum abnormality sub-threshold is marked as a vibration abnormal sample point.

8. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 7, characterized in that, The method for identifying temperature abnormal sample points in the real-time temperature gradient time series is as follows: Obtain the real-time temperature gradient abnormality degree sequence of the real-time temperature gradient time series, traverse the real-time temperature gradient abnormality degree sequence, and determine whether each temperature sample point in the real-time temperature gradient abnormality degree sequence exceeds the calibrated sub-threshold of the temperature gradient abnormality degree. If so, then the temperature sample points that exceed the calibrated sub-threshold of the temperature gradient abnormality degree are marked as temperature abnormal sample points.

9. A method for sensing the service performance of the meshing surface of a large-scale heavy-duty transmission gear according to claim 8, characterized in that The obtaining of the real-time temperature gradient anomaly degree sequence of the real-time temperature gradient time series includes: Calculating the statistical features of the real-time temperature gradient time series to form a real-time temperature gradient feature set; Using the principal component analysis method to reduce the dimension of the real-time temperature gradient feature set to obtain the low-dimensional real-time temperature gradient features; According to the low-dimensional real-time temperature gradient features and the support vector data description detection model, a real-time temperature gradient anomaly degree sequence is obtained.

10. A service performance sensing system for the meshing surface of a large heavy-duty transmission gear, which is used to implement a service performance sensing method for the meshing surface of a large heavy-duty transmission gear according to any one of claims 1-9, characterized in that, The system includes: An evaluation model construction module: used to obtain the first data of the target gear and establish a gear meshing surface service performance evaluation model according to the first data; the first data includes the historical meshing surface conductivity time series, historical vibration spectrum time series, and historical temperature gradient time series of the target gear under different working conditions; A service performance evaluation module: used to collect the real-time meshing surface conductivity time series, real-time vibration spectrum time series, and real-time temperature gradient time series of the target gear, and generate second data; according to the second data and the gear meshing surface service performance evaluation model, obtain the current service performance level of the target gear; A lubricating oil pollution state determination module: used to obtain the lubricating oil pollution state threshold according to the first data, dynamically calibrate the lubricating oil pollution state threshold according to the current service performance level of the target gear to obtain the calibrated lubricating oil pollution state threshold; judge whether the second data exceeds the calibrated lubricating oil pollution state threshold, if not, continue to monitor in real time; if it exceeds, perform online filtration of the lubricating oil to remove pollutants.

Images