Reduction gear operation state detection method based on big data

Abstract

The application relates to the technical field of speed reducer detection, and specifically discloses a speed reducer operation state detection method based on big data, which comprises the following steps: acquiring real-time operation data and historical operation data of a speed reducer, and acquiring historical vibration data and historical oil data according to the historical operation data; acquiring order spectrum characteristics according to the historical vibration data, and generating vibration characteristic benchmarks in a historical operation state based on the statistical distribution of the order spectrum characteristics; acquiring an oil particle concentration trend according to the historical oil data, and generating oil characteristic benchmarks in a conventional operation state based on the oil particle concentration trend. Through order tracking and equi-angle resampling technology, time-domain vibration signals are converted into angular-domain vibration signals, and order spectrum characteristics are extracted, so that the frequency ambiguity problem of a traditional frequency-domain analysis method under variable speed working conditions is effectively solved, and the accuracy of vibration characteristic extraction is improved.

Description

Technical Field

[0001] This application relates to the field of speed reducer testing technology, and in particular to a method for detecting the operating status of speed reducers based on big data. Background Technology

[0002] Gear reducers are core components of industrial transmission systems and are widely used in wind power generation, metallurgy, mining, port machinery, and other fields. The operating status of a gear reducer directly affects the safety and reliability of the entire production line. Once a malfunction occurs, it may lead to equipment downtime, production interruption, or even safety accidents, causing significant economic losses.

[0003] Currently, the condition detection methods for speed reducers mainly rely on a single signal source for analysis, and existing technologies have the following problems: Vibration analysis methods have poor adaptability under variable speed conditions. Traditional frequency domain analysis methods assume a constant equipment speed. When the reducer operates under variable speed and load conditions, frequency ambiguity occurs in the spectrum, leading to inaccurate feature extraction and difficulty in establishing reliable vibration characteristic benchmarks. Furthermore, oil analysis methods lack refined identification of wear types. Existing technologies typically judge the degree of wear based solely on the total amount of particles, failing to combine the distribution characteristics of particle composition (ferromagnetic and non-ferromagnetic) to identify wear types, resulting in insufficient ability to distinguish between different wear mechanisms such as abrasive wear, fatigue wear, and corrosive wear. The fusion analysis of vibration and oil signals is insufficient. Existing technologies typically treat vibration analysis and oil analysis as independent detection methods, lacking an effective multi-source information fusion mechanism. While some methods use simple weighted summation for fusion, the weights are fixed, failing to consider the differences in the equipment's operating stage or the data quality differences between different signal sources, leading to insufficient accuracy in the comprehensive evaluation results. Summary of the Invention

[0004] This application provides a method for detecting the operating status of a speed reducer based on big data, which at least solves the problem that judging the degree of wear solely based on the total amount of particles fails to combine the distribution characteristics of particle composition (ferromagnetic and non-ferromagnetic) to identify the type of wear, resulting in insufficient ability to distinguish between different wear mechanisms such as abrasive wear, fatigue wear, and corrosive wear.

[0005] Firstly, this application provides a method for detecting the operating status of a speed reducer based on big data, the method comprising: Acquire real-time and historical operating data of the reducer, and obtain historical vibration and historical oil data based on the historical operating data; The order spectrum features are obtained from the historical vibration data, and the vibration feature benchmark under historical operating conditions is generated based on the statistical distribution of the order spectrum features. The oil particle concentration trend is obtained based on the historical oil data, and an oil characteristic benchmark under normal operating conditions is generated based on the oil particle concentration trend. Real-time vibration data and real-time oil data are obtained based on the real-time operating data. Vibration state deviation value is obtained based on the real-time vibration data and vibration characteristic benchmark. Oil state deviation value is obtained based on the real-time oil data and oil characteristic benchmark. A comprehensive state evaluation value is generated based on the vibration state deviation value and the oil state deviation value, and it is determined whether the comprehensive state evaluation value is greater than a preset threshold. If the comprehensive status evaluation value is less than or equal to the preset threshold, the reducer is determined to be in normal operating condition. If the comprehensive status assessment value is greater than the preset threshold, the reducer is determined to be in an abnormal operating state.

[0006] Optionally, the step of obtaining the order spectrum characteristics based on the historical vibration data includes: Multiple sets of rotational speed pulse signals and vibration acceleration signals are obtained based on the historical vibration data. Instantaneous rotational speed is calculated based on the rotational speed pulse signals. The historical vibration data is then segmented based on the fluctuation rate of the instantaneous rotational speed to obtain several vibration data segments. The rotational speed fluctuation rate is obtained from the vibration data segment, and the vibration data segment with the rotational speed fluctuation rate less than the preset fluctuation threshold is determined as the steady-state vibration data segment. For each steady-state vibration data segment, order tracking is performed, and the steady-state vibration data segment is resampled at equal angles according to the rotational speed pulse signal to obtain the angular domain vibration signal; The angular domain vibration signal is subjected to a fast Fourier transform to obtain an order spectrum, and an amplitude sequence within a preset order range is extracted from the order spectrum. The coefficient of variation of the amplitude sequence of the same order is obtained from the steady-state vibration data segment. The order amplitude with the coefficient of variation less than the preset variation threshold is determined as the effective order amplitude, and the effective order amplitude is used as the order spectrum feature.

[0007] Optionally, the step of generating a vibration characteristic benchmark under historical operating conditions based on the statistical distribution of the order spectrum characteristics includes: The transmission chain parameters of the reducer are obtained, the vibration excitation source corresponding to each order is identified according to the transmission chain parameters, and the order spectrum features are divided into at least two order feature subsets according to the type of the vibration excitation source. Obtain the operating parameters corresponding to the historical operating data, wherein the operating parameters include speed and load; divide the historical operating data into multiple operating intervals according to the operating parameters; Within each of the aforementioned operating condition intervals, the mean and standard deviation of the amplitude of each order in each of the aforementioned order feature subsets are calculated respectively; Based on the mean and the standard deviation, determine the normal fluctuation range of each order amplitude within each operating condition range; Calculate the ratio of adjacent order amplitudes in each of the order feature subsets, and statistically analyze the distribution range of the ratios to determine the order correlation range of each of the order feature subsets; The normal fluctuation range of each order amplitude within each operating condition interval and the order correlation range of each order feature subset are used together as the vibration characteristic benchmark.

[0008] Optionally, the step of obtaining the oil particle concentration trend based on the historical oil data includes: Obtain particle count and particle composition data from the historical oil data, wherein the particle count includes the number of particles in multiple preset size ranges, and the particle composition data includes the content of ferromagnetic particles and the content of non-ferromagnetic particles; The ratio of ferromagnetic particles to non-ferromagnetic particles is calculated based on the particle composition data, and the wear type is identified based on the ratio, wherein the wear type includes abrasive wear, fatigue wear, and corrosive wear. The historical oil data is classified according to the wear type to obtain the oil data corresponding to each wear type; Sliding window sampling was performed on the oil data corresponding to each of the wear types, and the slope of the change in the number of particles in each size interval within each sampling window was calculated. Arrange the slopes of change in chronological order to obtain the particle concentration change trends for each size range under each wear type, and use these particle concentration change trends as the oil particle concentration trends.

[0009] Optionally, the step of generating an oil characteristic benchmark under normal operating conditions based on the oil particle concentration trend includes: The equipment runtime corresponding to the oil particle concentration trend is obtained, and the oil particle concentration trend is divided into the break-in stage, the stabilization stage and the degradation stage based on the equipment runtime. The median and interquartile range of particle concentration for each size range are obtained based on the oil particle concentration trends corresponding to the break-in stage, stabilization stage, and degradation stage. The upper and lower limits of particle concentration for each size range are obtained based on the median and the interquartile range. The ratio change is obtained based on the particle concentration in adjacent size ranges, and the ratio change is subjected to time-series smoothing to obtain a size distribution stability index. Based on the break-in stage, stabilization stage and degradation stage, the frequency percentage of each wear type is obtained, and the wear type with a frequency percentage greater than a preset percentage threshold is determined as the dominant wear type. The upper and lower limits of the reference, the size distribution stability index, and the dominant wear type of each size range are used together as the oil characteristic reference.

[0010] Optionally, the step of obtaining the vibration state deviation value based on real-time vibration data and vibration characteristic benchmarks includes: Obtain the current operating condition parameters corresponding to the real-time vibration data, and match the corresponding operating condition interval from the vibration characteristic benchmark according to the current operating condition parameters to obtain the target benchmark; The rotational speed pulse signal and vibration acceleration signal in the real-time vibration data are acquired, and the real-time vibration data is subjected to order tracking and equal angle resampling to obtain the real-time angular domain vibration signal. The real-time angular domain vibration signal is subjected to fast Fourier transform to extract real-time order spectrum features, and the real-time order spectrum features are grouped by order according to the order correlation range in the target reference. Calculate the degree of deviation between the amplitude of each order within each order group and the normal fluctuation range of the corresponding order in the target benchmark; The coupling deviation of the degree of deviation between adjacent orders within each order group is calculated based on the order association range, wherein the coupling deviation is used to reflect the degree of cooperative anomaly of the deviation between adjacent orders. The deviation degree and coupling deviation of each order group are weighted and summarized according to a preset weight to obtain the vibration state deviation value.

[0011] Optionally, the step of obtaining the oil state deviation value based on the real-time oil data and oil characteristic benchmark includes: Obtain the current equipment runtime corresponding to the real-time oil data, and match the upper limit, lower limit, size distribution stability index and dominant wear type of the corresponding stage from the oil feature benchmark; Obtain the real-time particle count and real-time particle composition data from the real-time oil data, and obtain the amount by which the real-time particle concentration in each size range exceeds the corresponding upper limit of the oil characteristic benchmark based on the real-time particle count and real-time particle composition data. The real-time content ratio of ferromagnetic particles to non-ferromagnetic particles is obtained based on the real-time particle composition data, and the current wear type is identified based on the real-time content ratio. Determine whether the current wear type is consistent with the dominant wear type of the corresponding stage. If they are inconsistent, set the wear type deviation to a preset deviation value; if they are consistent, the wear type deviation is zero. The ratio change of adjacent size intervals in the real-time particle concentration is obtained, and the ratio change is compared with the size distribution stability index in the oil characteristic benchmark to obtain the size distribution deviation. After normalizing the excess amount, wear type deviation amount, and size distribution deviation amount of each size range, the values ​​are weighted and summarized according to preset weights to obtain the oil state deviation value.

[0012] Optionally, the step of obtaining the real-time content ratio of ferromagnetic particles to non-ferromagnetic particles based on the real-time particle composition data, and identifying the current wear type based on the real-time content ratio, includes: Based on the real-time particle composition data, the ferromagnetic ratio of the number of ferromagnetic particles in each of the multiple preset size ranges is calculated to the total number of particles in that size range. Based on the ferromagnetic ratio of each size interval, the change in the ferromagnetic ratio between adjacent size intervals is calculated to obtain a gradient change sequence. The ferromagnetic proportion and gradient change sequence of each size range are compared with preset wear type determination conditions to determine the current wear type. The wear type determination conditions include the ferromagnetic proportion range corresponding to abrasive wear, the gradient change range corresponding to fatigue wear, and the non-ferromagnetic particle proportion threshold corresponding to corrosion wear.

[0013] Optionally, the step of generating a comprehensive condition assessment value based on the vibration state deviation value and the oil state deviation value includes: The current device runtime corresponding to the real-time operating data is obtained, and the current operating stage of the reducer is determined based on the current device runtime, wherein the operating stage includes the break-in stage, the stabilization stage and the degradation stage; The vibration reference weight and the oil reference weight are determined according to the operating stage, wherein, during the break-in stage, the oil reference weight is greater than the vibration reference weight; during the stabilization stage, the vibration reference weight is greater than the oil reference weight; and during the degradation stage, the vibration reference weight and the oil reference weight tend to be balanced. Obtain the vibration data confidence level of the real-time vibration data and the oil data confidence level of the real-time oil data; Based on the confidence levels of the vibration data and the oil data, the vibration reference weights and the oil reference weights are corrected respectively to obtain the final vibration weights and the final oil weights. The vibration state deviation value is weighted by the final vibration weight, the oil state deviation value is weighted by the final oil weight, and the weighted vibration state deviation value is added to the weighted oil state deviation value to obtain a weighted comprehensive value. Determine whether the directions of change of the vibration state deviation value and the oil state deviation value are opposite; If the directions of change are the same, then the weighted composite value is determined as the comprehensive state evaluation value; If the directions of change are opposite, the absolute value of the difference between the vibration state deviation value and the oil state deviation value is calculated as the cooperative deviation amount, and the weighted comprehensive value is added to the cooperative deviation amount to obtain the comprehensive state evaluation value.

[0014] Optionally, the step of determining whether the direction of change of the vibration state deviation value and the oil state deviation value is opposite includes: Obtain the vibration state deviation sequence and oil state deviation sequence obtained from the most recent preset number of state assessment calculations; Linear regression was performed on the vibration state deviation value sequence and the oil state deviation value sequence respectively to calculate the vibration deviation slope and the oil deviation slope; If the product of the vibration deviation slope and the oil deviation slope is negative, the direction of change is determined to be opposite; otherwise, the direction of change is determined to be the same.

[0015] Compared with related technologies, this application has at least the following technical effects: By employing order tracking and equal-angle resampling techniques, time-domain vibration signals are converted into angular-domain vibration signals, and order spectrum features are extracted. This effectively solves the frequency ambiguity problem of traditional frequency domain analysis methods under variable-speed conditions, improving the accuracy of vibration feature extraction. By establishing vibration feature benchmarks for different operating conditions, and setting the normal fluctuation range and order correlation range for each order amplitude under different speeds and loads, refined characterization of vibration features under different operating conditions is achieved, reducing the false alarm rate caused by changes in operating conditions. Through a wear type identification method based on the ratio of ferromagnetic to non-ferromagnetic particle content and its gradient change sequence, refined differentiation of abrasive wear, fatigue wear, and corrosive wear is achieved, enhancing the ability of oil analysis to identify wear mechanisms. By dividing the oil particle concentration trend into break-in, stable, and degradation stages, and establishing oil feature benchmarks for each stage, the oil detection standards can adapt to different stages of the equipment lifecycle, improving the targeting and accuracy of detection. Through an adaptive weight allocation mechanism during operation, the fusion weights of vibration and oil signals are dynamically adjusted according to the equipment's break-in, stabilization, or degradation phase, avoiding the poor adaptability of fixed weights at different stages. By introducing vibration and oil data confidence levels to correct the baseline weights, dynamic weight adjustment based on data quality is achieved, reducing the interference of low-quality data on the comprehensive evaluation results. Through a collaborative deviation detection mechanism, when the deviation values ​​of vibration and oil states change in opposite directions, an additional collaborative deviation is added, improving the ability to detect contradictory abnormal signs and reducing the risk of missed detections.

[0016] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is an overall flowchart illustrating a method for detecting the operating status of a speed reducer based on big data, according to an exemplary embodiment.

[0018] Figure 2 This is a flowchart illustrating the acquisition of order spectrum features according to an exemplary embodiment.

[0019] Figure 3 This is a flowchart illustrating the generation of vibration characteristic references according to an exemplary embodiment.

[0020] Figure 4 This is a flowchart illustrating the trend of oil particle concentration according to an exemplary embodiment.

[0021] Figure 5 This is a flowchart illustrating the generation of oil feature references according to an exemplary embodiment.

[0022] Figure 6 This is a flowchart illustrating the acquisition of vibration state deviation values ​​according to an exemplary embodiment.

[0023] Figure 7 This is a flowchart illustrating the acquisition of oil state deviation values ​​according to an exemplary embodiment.

[0024] Figure 8 This is a flowchart illustrating the identification of the current wear type according to an exemplary embodiment.

[0025] Figure 9 This is a flowchart illustrating the generation of a comprehensive state assessment value according to an exemplary embodiment. Figure 10 This is a flowchart illustrating the determination of the direction of change of vibration state deviation value and oil state deviation value according to an exemplary embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0027] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0028] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0029] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0030] In related technologies, the main shortcomings of the gearbox condition detection methods are as follows: First, vibration analysis methods lack adaptability to various operating conditions. Traditional frequency domain analysis methods rely on Fast Fourier Transform (FFT) to perform spectral analysis on vibration signals, assuming the signal is stationary, i.e., the equipment speed is constant. However, in actual engineering, reducers often operate under varying speed and load conditions. In this case, the vibration signal becomes non-stationary, and the traditional FFT method produces frequency ambiguity, causing broadening and aliasing of frequency components in the spectrum, making it impossible to accurately extract vibration characteristics of each order. Furthermore, traditional methods typically use a uniform fixed threshold for anomaly detection across all operating conditions, failing to consider the normal fluctuations in vibration characteristics under different conditions, which can easily lead to false alarms during operating condition switching.

[0031] Secondly, oil analysis methods lack the ability to identify wear types. Existing oil detection technologies mainly focus on the total amount of particles and particle size distribution, judging the degree of wear by setting particle concentration thresholds. However, relying solely on the total amount of particles cannot distinguish different wear mechanisms. The particles generated by abrasive wear, fatigue wear, and corrosive wear differ significantly in composition (ratio of ferromagnetic to non-ferromagnetic) and size distribution, and existing technologies fail to fully utilize these differences for refined identification of wear types.

[0032] Third, the multi-source signal fusion mechanism is imperfect. Vibration and hydraulic signals from the reducer reflect the equipment's operating status from different perspectives: vibration signals mainly reflect the dynamic response of the mechanical structure, while hydraulic signals mainly reflect the wear state of the friction pairs. Fusion analysis of both can improve the comprehensiveness and accuracy of fault detection. However, current technologies typically analyze both independently and then simply weight and sum them, without considering the impact of the equipment's operating phase on the indicative meaning of the two signals, nor the impact of data quality differences on the fusion results, leading to insufficient accuracy in the comprehensive assessment.

[0033] Based on the above, embodiments of the present invention provide a method for detecting the operating status of a speed reducer based on big data, which will be described in detail below with reference to specific embodiments and accompanying drawings.

[0034] Example 1

[0035] This invention provides a method for detecting the operating status of a speed reducer based on big data. Figure 1 This is an overall flowchart illustrating a method for detecting the operating status of a speed reducer based on big data, according to an exemplary embodiment. Figure 1 As shown, the method includes the following steps: S1. Obtain real-time and historical operating data of the reducer, and obtain historical vibration data and historical oil data based on the historical operating data; In this embodiment, the reducer's operating data includes vibration signals (including speed pulse signals and vibration acceleration signals) collected by vibration sensors, oil data (including particle count values ​​and particle composition data) collected by oil sensors, and operating parameters (including speed and load). Historical operating data refers to the operating data accumulated by the reducer over a past period, used to establish characteristic benchmarks; real-time operating data refers to the operating data collected at the current moment, used for condition assessment. Specifically, historical vibration data is extracted from vibration sensor data in historical operating data, and historical oil data is extracted from oil sensor data in historical operating data. S2. Obtain the order spectrum characteristics based on the historical vibration data, and generate the vibration characteristic benchmark under the historical operating state based on the statistical distribution of the order spectrum characteristics. In this embodiment, the order spectrum feature refers to the feature obtained by performing spectral analysis on the vibration signal in the angular domain (rather than the time domain). It uses multiples of the rotational speed (order) as the frequency axis, which can eliminate the influence of rotational speed changes on the spectrum. The vibration feature benchmark refers to the reference range that characterizes the vibration amplitude of each order of the reducer under normal operating conditions, established based on a large amount of historical data statistics.

[0036] S3. Obtain the oil particle concentration trend based on the historical oil data, and generate an oil characteristic benchmark under normal operating conditions based on the oil particle concentration trend. In this embodiment, the oil particle concentration trend refers to the change in particle concentration over time for each size range. The oil characteristic benchmark refers to a reference standard established based on historical oil data to characterize the wear particle characteristics of the reducer under normal operating conditions.

[0037] S4. Obtain real-time vibration data and real-time oil data based on the real-time operating data, obtain vibration state deviation value based on the real-time vibration data and vibration characteristic benchmark, and obtain oil state deviation value based on the real-time oil data and oil characteristic benchmark. In this embodiment, the vibration state deviation value is a comprehensive quantitative index that reflects the degree of deviation of real-time vibration data from the vibration characteristic benchmark. The larger the deviation value, the more abnormal the vibration state of the equipment.

[0038] S5. Generate a comprehensive state evaluation value based on the vibration state deviation value and the oil state deviation value, and determine whether the comprehensive state evaluation value is greater than a preset threshold. If the comprehensive status evaluation value is less than or equal to the preset threshold, the reducer is determined to be in normal operating condition. If the comprehensive status assessment value is greater than the preset threshold, the reducer is determined to be in an abnormal operating state.

[0039] In this embodiment, the preset threshold is determined based on the distribution of comprehensive status evaluation values ​​of normal operation and fault status in historical data. The threshold that optimizes detection accuracy and false alarm rate can be selected using ROC curve analysis. When the comprehensive status evaluation value exceeds this threshold, the system triggers an abnormal alarm, reminding maintenance personnel to inspect or maintain the reducer.

[0040] Specifically, such as Figure 2 As shown, the steps in step S2 for obtaining the order spectral features include: S21. Obtain multiple sets of rotational speed pulse signals and vibration acceleration signals based on the historical vibration data, calculate the instantaneous rotational speed based on the rotational speed pulse signals, and segment the historical vibration data according to the fluctuation rate of the instantaneous rotational speed to obtain several vibration data segments.

[0041] In this embodiment, the speed pulse signal is generated by a photoelectric encoder or magnetoelectric sensor installed on the input or output shaft of the reducer. Each pulse corresponds to a fixed angle rotated by the shaft. The instantaneous speed can be calculated by measuring the time interval between adjacent pulses. The fluctuation rate of the instantaneous speed is defined as the ratio of the difference between adjacent instantaneous speeds to the average speed, reflecting the stability of the speed. When the fluctuation rate is large, it indicates that the equipment is in a transient process of acceleration, deceleration, or sudden load change. The vibration data at this time is not suitable for establishing a steady-state characteristic benchmark.

[0042] S22. Obtain the rotational speed fluctuation rate based on the vibration data segment, and determine the vibration data segment with the rotational speed fluctuation rate less than the preset fluctuation threshold as the steady-state vibration data segment.

[0043] In this embodiment, the preset fluctuation threshold is set based on the rated speed of the reducer and the allowable speed fluctuation range. For example, the preset fluctuation threshold is determined based on the statistical characteristics of the actual operating data of the reducer. The specific method is as follows: The speed signal of the reducer under stable operating conditions (constant load, no acceleration or deceleration) for at least 30 seconds is collected, and the 95th quantile of the instantaneous speed fluctuation rate is calculated. 1.2 times this quantile is used as the preset fluctuation threshold. For reference, for most industrial reducers, the threshold obtained by this statistical method usually falls within the range of 2% to 5% of the rated speed; for high-precision reducers (such as those used in precision machine tools), the threshold may be close to the lower limit, and for general industrial reducers (such as those used in conveyors), the threshold may be close to the upper limit, but the actual threshold should be based on the statistical calculation results. By screening steady-state vibration data segments, the interference of acceleration and deceleration transient processes on feature extraction is eliminated, ensuring the accuracy of subsequent order analysis.

[0044] S23. Perform order tracking on each of the steady-state vibration data segments, and resample the steady-state vibration data segments at equal angles according to the rotational speed pulse signal to obtain the angular domain vibration signal.

[0045] In this embodiment, order tracking is a technique for converting a non-stationary signal in the time domain into a stationary signal in the angular domain. The vibration acceleration signal is resampled at equal angular intervals using the rotation angle of the shaft as the independent variable (rather than time). Specifically, the time points at which the shaft rotates through a fixed angle (e.g., 1 degree or 0.5 degrees) are determined based on the rotational speed pulse signal. At these time points, the original vibration acceleration signal is interpolated and sampled to obtain the angular domain vibration signal. After equal angular resampling, the vibration order components, whose frequency varied with rotational speed in the time domain, become fixed orders in the angular domain, thus eliminating the influence of rotational speed variations on the frequency spectrum.

[0046] S24. Perform a fast Fourier transform on the angular domain vibration signal to obtain the order spectrum, and extract the amplitude sequence within a preset order range from the order spectrum.

[0047] In this embodiment, after performing an FFT transform on the angular domain vibration signal, the horizontal axis represents the order (i.e., a multiple of the rotational speed), and the vertical axis represents the vibration amplitude of the corresponding order. The preset order range is determined based on the transmission chain parameters of the reducer and the vibration excitation source of interest; for example, it can be set from order 1 to order 50. The amplitude corresponding to each order is extracted from the order spectrum to form an amplitude sequence.

[0048] S25. Obtain the coefficient of variation of the amplitude sequence of the same order based on the steady-state vibration data segment, determine the order amplitude with the coefficient of variation less than the preset variation threshold as the effective order amplitude, and use the effective order amplitude as the order spectrum feature. ; In the formula, represents the coefficient of variation for the j-th order, used to measure the dispersion of the amplitude at that order, and is dimensionless. Here, j is the order number (a positive integer, such as 1st order, 2nd order, etc.). This represents the standard deviation of the amplitude of the j-th order vibration across all steady-state vibration data segments, with the same unit as the amplitude (e.g., m / s). 2 ). This represents the mean amplitude of the j-th order vibration across all steady-state vibration data segments, with the same unit as the amplitude. When If the amplitude is less than the preset variation threshold (usually 0.3), the order amplitude is determined to be a valid order amplitude; otherwise, it is excluded.

[0049] In this embodiment, the coefficient of variation is defined as the ratio of the standard deviation to the mean, reflecting the dispersion of the data. The coefficient of variation is calculated for the amplitude of the same order across multiple steady-state vibration data segments. Orders with smaller coefficients of variation indicate better consistency and repeatability of their amplitudes across different time periods, making them suitable as stable characteristic indicators. The preset variation threshold is set based on engineering experience, for example, ranging from 0.2 to 0.4. A typical value is 0.3. The method for determining this threshold is as follows: historical vibration data is collected multiple times for the same reducer under the same operating conditions, and the coefficient of variation for each order amplitude is calculated. The median or 75th percentile of the coefficient of variation distribution is selected as the threshold. If the coefficient of variation is greater than 0.3, it indicates that the amplitude of that order fluctuates too much and is easily affected by noise or non-steady-state factors, making it unsuitable as a feature. Order amplitudes with coefficients of variation greater than the preset variation threshold may be affected by noise or other random factors and are excluded.

[0050] like Figure 3 As shown, step S2, which generates a vibration characteristic reference based on the statistical distribution of the order spectrum features, includes: S26. Obtain the transmission chain parameters of the reducer, identify the vibration excitation source corresponding to each order according to the transmission chain parameters, and divide the order spectrum features into at least two order feature subsets according to the type of the vibration excitation source.

[0051] In this embodiment, the transmission chain parameters include the number of teeth on each gear, the speed ratio of the shaft, and the number of rolling elements in the bearing. Based on these parameters, the order of each gear's meshing frequency and the bearing's characteristic frequency can be calculated. For example, if the number of teeth on the driving gear of a certain stage is Z1 and the number of teeth on the driven gear is Z2, then the meshing order of that stage is Z1 (or Z2). According to the type of vibration excitation source (such as gear meshing, bearing rolling, shaft imbalance, etc.), the order spectrum features are divided into subsets of gear meshing order, bearing order, and shaft frequency order. This grouping method based on physical mechanisms allows subsequent anomaly detection to pinpoint the specific fault source.

[0052] S27. Obtain the operating parameters corresponding to the historical operating data, wherein the operating parameters include speed and load; divide the historical operating data into multiple operating intervals according to the operating parameters.

[0053] In this embodiment, the operating parameters include the input speed and output load of the reducer. The speed and load are each divided into several intervals (for example, the speed is divided into low speed, medium speed, and high speed intervals, and the load is divided into light load, medium load, and heavy load intervals), which are combined to form multiple operating condition intervals (such as low speed light load, medium speed medium load, etc.). Each operating condition interval corresponds to a set of historical operating data.

[0054] S28. Within each of the aforementioned operating condition intervals, calculate the mean and standard deviation of the amplitude of each order in each of the aforementioned order feature subsets.

[0055] S29. Based on the mean and the standard deviation, determine the normal fluctuation range of each order amplitude within each operating condition interval.

[0056] ; In the formula, This represents the maximum value within the normal fluctuation range. This represents the minimum value within the normal fluctuation range. This represents the mean value of the j-th order amplitude within the operating condition interval k, with the same unit as the amplitude. This represents the standard deviation of the j-th order amplitude within the operating condition interval k, with the same unit. k is the operating condition interval number, determined by the combination of the speed and load of the reducer (a positive integer).

[0057] In this embodiment, the normal fluctuation range can be set as the mean plus or minus a certain number of standard deviations, for example, the mean ± 3σ, where σ is the standard deviation. This range covers approximately 99.7% of normal data points, and amplitudes exceeding this range are considered potentially abnormal.

[0058] S210. Calculate the ratio of adjacent order amplitudes in each of the order feature subsets, and statistically analyze the distribution range of the ratios to determine the order correlation range of each of the order feature subsets.

[0059] In this embodiment, the ratio of adjacent order amplitudes reflects the energy distribution relationship between orders. For example, in the gear meshing order subset, the amplitude ratio between the meshing frequency and its harmonics (2nd meshing frequency, 3rd meshing frequency, etc.) usually remains within a certain range under normal conditions. When gears experience wear or pitting, the relative amplitudes of the harmonic components change, causing the order ratio to deviate from the normal range. The distribution range of each order ratio in historical statistical data (e.g., the 5th to 95th percentile range) is used as the order correlation range.

[0060] S211. The normal fluctuation range of each order amplitude within each operating condition interval and the order correlation range of each order feature subset are used together as the vibration characteristic benchmark.

[0061] Specifically, such as Figure 4 As shown, step S3, which involves obtaining the oil particle concentration trend, includes: S31. Obtain the particle count value and particle composition data from the historical oil data, wherein the particle count value includes the number of particles in multiple preset size ranges, and the particle composition data includes the content of ferromagnetic particles and the content of non-ferromagnetic particles.

[0062] In this embodiment, particle count values ​​are obtained through online particle counters or offline oil analysis, and are divided into multiple ranges according to particle size, such as 4-6μm, 6-14μm, 14-21μm, 21-38μm, 38-70μm, and 70μm. Particle composition data are obtained through spectral analysis or ferrographic analysis to distinguish between ferromagnetic particles (mainly from steel wear of gears and bearings) and non-ferromagnetic particles (mainly from wear of copper alloy bushings, seals, etc.).

[0063] S32. Calculate the content ratio of ferromagnetic particles to non-ferromagnetic particles based on the particle composition data, and identify the wear type based on the content ratio, wherein the wear type includes abrasive wear, fatigue wear and corrosive wear.

[0064] In this embodiment, the ratio of ferromagnetic particles to non-ferromagnetic particles is an important indicator for distinguishing wear types. A high ratio indicates that ferromagnetic particles are dominant, which usually corresponds to abrasive wear or fatigue wear; a low ratio indicates that non-ferromagnetic particles account for a higher proportion, which may correspond to corrosion wear or abnormal wear of copper alloy components.

[0065] S33. Classify the historical oil data according to the wear type to obtain oil data corresponding to each type of wear.

[0066] S34. Perform sliding window sampling on the oil data corresponding to each of the wear types, and calculate the slope of the change in the number of particles in each size interval within each sampling window.

[0067] In this embodiment, the width of the sliding window is set according to the oil sampling frequency and the equipment operating cycle; for example, it can be set to five consecutive samples. The slope of change is calculated using linear regression or a difference method, reflecting the rate of change of particle concentration within this time window.

[0068] S35. Arrange the slope of change in chronological order to obtain the particle concentration change trend of each size range under each wear type, and take the particle concentration change trend as the oil particle concentration trend.

[0069] like Figure 5 As shown, step S3, which generates an oil characteristic benchmark based on the oil particle concentration trend, includes: S36. Obtain the equipment running time corresponding to the oil particle concentration trend, and divide the oil particle concentration trend into the running-in stage, the stabilization stage and the degradation stage according to the equipment running time.

[0070] In this embodiment, the break-in stage refers to the initial operation of the equipment (e.g., the first 500 operating hours). During this stage, the friction pairs are in a break-in state, and the particle concentration is high and changes rapidly. The stable stage refers to the normal operation stage of the equipment (e.g., 500 to 5000 operating hours). During this stage, the wear rate tends to stabilize, and the particle concentration remains at a low level. The degradation stage refers to the aging stage of the equipment (e.g., after 5000 operating hours). During this stage, wear intensifies, and the particle concentration shows an upward trend. The division of each stage can be automatically identified based on the inflection point of the particle concentration trend, or it can be preset according to the maintenance manual provided by the equipment manufacturer.

[0071] S37. Based on the oil particle concentration trends corresponding to the break-in stage, stabilization stage, and degradation stage, obtain the median and interquartile range of particle concentration for each size range.

[0072] In this embodiment, the median and interquartile range are robust statistics, unaffected by extreme values, and suitable for describing the distribution characteristics of particle concentration. The interquartile range (IQR) is defined as the difference between the third quartile and the first quartile, reflecting the dispersion of the data.

[0073] S38. Obtain the baseline upper and lower limits of particle concentration for each size range based on the median and the interquartile range.

[0074] In this embodiment, the upper limit of the reference range can be set to the median plus a certain multiple of the IQR, and the lower limit of the reference range can be set to the median minus a certain multiple of the IQR, but not less than zero. A particle concentration exceeding the upper limit of the reference range indicates an abnormally high wear rate, while a particle concentration below the lower limit of the reference range may indicate abnormal sampling or the initial state after oil replacement.

[0075] S39. Obtain the ratio change based on the particle concentration in adjacent size intervals, and perform time-series smoothing on the ratio change to obtain the size distribution stability index.

[0076] In this embodiment, the change in the ratio of particle concentration between adjacent size ranges reflects the stability of the particle size distribution. Under normal wear conditions, the particle concentration in each size range maintains a relatively stable proportional relationship; when the wear mechanism changes (such as from abrasive wear to fatigue wear), the proportion of large-sized particles increases significantly, leading to a larger change in the ratio. Time-series smoothing uses moving average or exponential smoothing methods to eliminate the impact of short-term fluctuations.

[0077] S310. Based on the break-in stage, stabilization stage and degradation stage, obtain the frequency ratio of each wear type, and determine the wear type with the frequency ratio greater than the preset ratio threshold as the dominant wear type.

[0078] In this embodiment, the dominant wear type reflects the most common wear mechanism in each operating stage of the equipment. For example, the dominant wear type in the break-in stage is usually abrasive wear, the dominant wear type in the stable stage is usually fatigue wear, and multiple wear types may coexist in the degradation stage. A preset percentage threshold is used to determine whether a certain wear type belongs to the dominant type in that operating stage. For example, the preset percentage threshold ranges from 50% to 70%, with a typical value of 60%. The determination method is as follows: statistically analyze the frequency of each type of wear in historical oil data, and based on the percentage corresponding to the most frequent type, set 80% to 90% of this percentage as the threshold, or directly use the empirical value of 60%. If the percentages of multiple wear types all exceed the threshold, the type with the highest percentage is taken as the dominant type.

[0079] S311. The upper limit, lower limit, size distribution stability index, and dominant wear type of each size range are used together as the oil characteristic reference.

[0080] Specifically, such as Figure 6 As shown, step S4 includes: S41. Obtain the current operating condition parameters corresponding to the real-time vibration data, and match the corresponding operating condition interval from the vibration characteristic reference according to the current operating condition parameters to obtain the target reference.

[0081] In this embodiment, the operating condition range of the equipment is determined based on the current rotational speed and load. The normal fluctuation range and order correlation range corresponding to this operating condition range are selected as the target benchmark from the vibration characteristic benchmark. This benchmark selection method based on operating condition matching ensures that the reference standard for anomaly detection is adapted to the current operating conditions.

[0082] S42. Obtain the rotational speed pulse signal and vibration acceleration signal from the real-time vibration data, perform order tracking and equal-angle resampling on the real-time vibration data, and obtain the real-time angular domain vibration signal.

[0083] S43. Perform a fast Fourier transform on the real-time angular domain vibration signal to extract the real-time order spectrum features, and group the real-time order spectrum features by order according to the order correlation range in the target reference.

[0084] In this embodiment, the amplitude of each order is extracted after performing FFT transformation on the real-time angular domain vibration signal, and each order is divided into gear meshing order group, bearing order group, etc. according to the order correlation range in the target reference.

[0085] S44. Calculate the degree of deviation between the amplitude of each order in each order group and the normal fluctuation range of the corresponding order in the target benchmark.

[0086] ; In the formula, It represents the degree of deviation of the j-th order within the operating condition interval k, is dimensionless, and takes a value greater than 0. This represents the amplitude of the j-th order in the real-time vibration data, with the same unit as the baseline mean. and These are the mean and standard deviation of the j-th order amplitude within the k-th operating condition interval in the target benchmark. When the order amplitude... Falling within the normal fluctuation range At that time, =0; when out of range It is the ratio of the excess to half the width 3σ.

[0087] In this embodiment, the deviation is calculated as follows: when the order amplitude is within the normal fluctuation range, the deviation is zero; when the order amplitude is within the normal fluctuation range... Within the time frame, the degree of deviation When the order amplitude is below the lower limit or above the upper limit, the deviation is the excess amount (i.e., ) and half width The ratio of .

[0088] S45. Calculate the coupling deviation of the degree of deviation between adjacent orders within each order group based on the order association range, wherein the coupling deviation is used to reflect the degree of cooperative anomaly of the deviation between adjacent orders.

[0089] ; In the formula, This represents the coupling deviation of the g-th order group, which is dimensionless and has a value ≥ 0. g is the order group number, which is grouped according to the type of vibration excitation source (such as gear meshing order group, bearing order group, etc.). This indicates the degree of deviation of the j-th order within the g-th order group. This indicates the degree of deviation of the (j+1)th order (adjacent orders) within the same order group; both have the same meaning as in step S44. Consistent. This represents the total number of orders (a positive integer) contained in the g-th order group. The coupling deviation occurs when multiple adjacent orders within the same order group deviate simultaneously. A significant increase reflects the degree of coordination anomaly; if there is only a single isolated deviation, then... .

[0090] In this embodiment, the physical meaning of coupling deviation is that when multiple adjacent orders within the same order group deviate simultaneously, the reliability of its indication of anomaly is much higher than that of an isolated deviation of a single order. For example, if the gear meshing frequency and its 2nd and 3rd harmonics are simultaneously elevated, it strongly suggests that the gear has wear or pitting faults. Coupling deviation is quantified by calculating the covariance or joint deviation probability of the degree of deviation of adjacent orders within the group.

[0091] S46. The deviation degree and the coupling deviation of each order group are weighted and summarized according to a preset weight to obtain the vibration state deviation value; ; In the formula, This represents the deviation from the vibration state, is dimensionless, and has a value ≥ 0. The deviation weight (preset constant) represents the degree of deviation of the g-th order group. The coupling bias weight (preset constant) represents the weight of the g-th order group, usually... G represents the total number of order groups (a positive integer). and The meaning is consistent with that in steps S44 and S45. The total number of orders within the g-th order group has the same meaning as in step S45. The formula sums the weighted average deviation and coupling deviation within each order group to obtain the overall vibration state deviation value.

[0092] In this embodiment, the preset weights are used to integrate deviation information from different order groups (such as gear meshing order group, bearing order group, and shaft frequency order group). For example, the sum of the weights of each order group is 1, and a typical allocation is: 0.5 for the gear meshing order group, 0.3 for the bearing order group, and 0.2 for the shaft frequency order group. The determination method is as follows: weights are assigned based on the historical failure rate of each component in the reducer transmission chain; the higher the failure rate, the greater the weight of the corresponding order group. If there are no historical failure statistics, the above typical values ​​are used. For the degree of deviation and coupling deviation within the same order group, the weight ratio is 0.6:0.4.

[0093] Specifically, such as Figure 7 As shown, step S4 further includes: S47. Obtain the current equipment running time corresponding to the real-time oil data, and match the upper limit of the reference, the lower limit of the reference, the size distribution stability index and the dominant wear type of the corresponding stage from the oil feature reference.

[0094] S48. Obtain the real-time particle count value and real-time particle composition data from the real-time oil data, and obtain the amount by which the real-time particle concentration in each size range exceeds the corresponding upper limit of the oil characteristic reference based on the real-time particle count value and real-time particle composition data.

[0095] In this embodiment, when the real-time particle concentration is lower than the upper limit of the reference range, the excess amount is zero; when the real-time particle concentration is higher than the upper limit of the reference range, the excess amount is the difference between the actual value and the upper limit of the reference range.

[0096] S49. Obtain the real-time content ratio of ferromagnetic particles to non-ferromagnetic particles based on the real-time particle composition data, and identify the current wear type based on the real-time content ratio.

[0097] S410. Determine whether the current wear type is consistent with the dominant wear type of the corresponding stage. If they are inconsistent, set the wear type deviation to a preset deviation value. If they are consistent, the wear type deviation is zero.

[0098] In this embodiment, a preset deviation value is used to quantify the abnormal contribution when the current wear type is inconsistent with the dominant wear type, and the value ranges from 0.3 to 0.8 (after normalization). A typical value is 0.5. Determination method: After normalizing the wear type deviation and other deviations (excess amount, size distribution deviation) to the [0,1] interval, the preset deviation value should reflect the severity of the wear type anomaly, and is usually set to a moderately high value of 0.5. It can also be adjusted according to the correlation between wear type changes and equipment failures in historical failure cases; the higher the correlation, the larger the value. When the wear type identified in real time is inconsistent with the dominant wear type at that stage, it indicates that the wear state of the equipment may have undergone an abnormal change, which requires attention. The preset deviation value is set according to the severity of the wear type change.

[0099] S411. Obtain the ratio change of adjacent size intervals in the real-time particle concentration, and compare the ratio change with the size distribution stability index in the oil characteristic benchmark to obtain the size distribution deviation.

[0100] S412. After normalizing the excess amount, wear type deviation amount and size distribution deviation amount of each size range, the oil state deviation value is obtained by weighting and summarizing according to the preset weight.

[0101] In this embodiment, as Figure 8 As shown, the steps to identify the current wear type include: S491. Based on the real-time particle composition data, calculate the ferromagnetic ratio of the number of ferromagnetic particles in each of the multiple preset size intervals to the total number of particles in that size interval.

[0102] S492. Based on the ferromagnetic ratio of each size interval, calculate the change in the ferromagnetic ratio between adjacent size intervals to obtain a gradient change sequence.

[0103] In this embodiment, the gradient change sequence reflects the distribution pattern of ferromagnetic particle content with particle size. Different wear types exhibit different characteristic patterns in the gradient change sequence: abrasive wear is usually characterized by a high proportion of ferromagnetic particles in all size ranges and a gentle gradient change; fatigue wear is usually characterized by a significant increase in the proportion of ferromagnetic particles in the large size range and a large gradient change; corrosion wear is usually characterized by a low proportion of ferromagnetic particles and a high proportion of non-ferromagnetic particles (such as oxides).

[0104] S493. The ferromagnetic ratio and gradient change sequence of each size range are compared with preset wear type determination conditions to determine the current wear type. The wear type determination conditions include the ferromagnetic ratio range corresponding to abrasive wear, the gradient change range corresponding to fatigue wear, and the non-ferromagnetic particle ratio threshold corresponding to corrosion wear.

[0105] In this embodiment, as Figure 9 As shown, step S5, which generates the comprehensive state assessment value, includes the following steps: S51. Obtain the current device runtime corresponding to the real-time operating data, and determine the current operating stage of the reducer based on the current device runtime, wherein the operating stage includes the break-in stage, the stabilization stage, and the degradation stage.

[0106] S52. Determine the vibration reference weight and the oil reference weight according to the operating stage, wherein, in the break-in stage, the oil reference weight is greater than the vibration reference weight; in the stabilization stage, the vibration reference weight is greater than the oil reference weight; and in the degradation stage, the vibration reference weight and the oil reference weight tend to be balanced.

[0107] In this embodiment, the physical basis for the adaptive weight allocation during the running phase is as follows: During the break-in phase, the friction pair is undergoing surface morphology shaping, and the wear particles in the oil change drastically and are rich in information. Meanwhile, the vibration signal may fluctuate significantly due to changes in surface roughness caused by break-in, thus the oil signal has a stronger indicative significance. During the stable phase, a stable lubricating film has formed on the friction pair surface, and the oil particle concentration remains at a low level and changes slowly. The vibration signal can more sensitively reflect early defects in gears or bearings (such as pitting, cracks, etc.), thus the vibration signal has a stronger indicative significance. During the degradation phase, increased wear leads to a significant increase in oil particle concentration, and obvious fault characteristics also appear in the vibration signal. Both signals have important indicative significance, therefore the weights tend to be balanced. For example, during the break-in phase, the oil reference weight is 0.6–0.8, and the vibration reference weight is 0.2–0.4; during the stable phase, the vibration reference weight is 0.6–0.8, and the oil reference weight is 0.2–0.4; during the degradation phase, both are 0.5. During the break-in period, the oil reference weight is 0.7, and during the stabilization period, the vibration reference weight is 0.7.

[0108] S53. Obtain the vibration data confidence level of the real-time vibration data and the oil data confidence level of the real-time oil data.

[0109] In this embodiment, the confidence level of vibration data is determined based on the signal-to-noise ratio (SNR) of the real-time vibration signal. A higher SNR indicates a larger proportion of effective components and less noise interference in the vibration signal, resulting in higher data quality and higher confidence. The confidence level of oil data is determined based on the effective particle count rate of real-time oil sampling. The effective particle count rate refers to the proportion of particles successfully identified and counted by the sensor to the total number of particles actually passing through the sensor. A higher proportion indicates higher reliability of the oil data.

[0110] S54. Based on the confidence levels of the vibration data and the oil data, the vibration reference weight and the oil reference weight are corrected respectively to obtain the final vibration weight and the final oil weight.

[0111] In this embodiment, the weight correction method is as follows: the final vibration weight is equal to the vibration reference weight multiplied by the vibration data confidence level, and the final oil weight is equal to the oil reference weight multiplied by the oil data confidence level. Then, the two are normalized to ensure that the sum of the final weights is 1. When the data quality of a certain signal source is poor, its confidence level is low, and the corresponding final weight is reduced accordingly, thereby reducing the interference of low-quality data on the comprehensive evaluation results.

[0112] S55. The vibration state deviation value is weighted by the final vibration weight, the oil state deviation value is weighted by the final oil weight, and the weighted vibration state deviation value is added to the weighted oil state deviation value to obtain a weighted comprehensive value.

[0113] ; In the formula, This represents the weighted composite value, which is dimensionless. and These represent the final weights of vibration and oil, respectively. This indicates the deviation value of the vibration state. This represents the deviation value of the oil condition, which is a weighted sum based on the excess amount of oil particle concentration, the deviation amount of wear type, and the deviation amount of size distribution. S56. Determine whether the directions of change of the vibration state deviation value and the oil state deviation value are opposite.

[0114] In this embodiment, opposite directions of change refer to a situation where the vibration state deviation value shows an upward trend while the oil state deviation value shows a downward trend, or vice versa. This indicates that two independent signal sources are giving contradictory abnormal indications, which is an abnormal sign that requires close attention. For example, the vibration signal shows that the gear meshing abnormality has intensified (vibration deviation value increases), but the particle concentration in the oil does not show a significant change (oil deviation value is stable or decreasing), which may suggest that the fault is still in its early stage and the oil signal has not yet reflected the abnormality; or conversely, the oil particle concentration increases sharply but the vibration signal is normal, which may suggest that the wear particles come from non-critical components or sampling abnormalities. The judgment of opposite directions of change is made using the sliding window linear regression slope method. Assuming the sliding window length L=5, the vibration state deviation value sequence and the oil state deviation value sequence calculated from the most recent 5 state assessments are collected, and their linear regression slopes are calculated respectively. If the product of the two slopes is negative, it indicates that one is on an upward trend and the other is on a downward trend, that is, the directions of change are opposite; otherwise, the directions are the same. This method can effectively suppress the instantaneous symbol fluctuations caused by sampling noise and improve the stability of trend judgment.

[0115] If the directions of change are the same, then the weighted composite value is determined as the comprehensive state evaluation value.

[0116] If the directions of change are opposite, the absolute value of the difference between the vibration state deviation value and the oil state deviation value is calculated as the cooperative deviation amount, and the weighted comprehensive value is added to the cooperative deviation amount to obtain the comprehensive state evaluation value.

[0117] ; In the formula, This represents the overall state assessment value, which is dimensionless. This represents the weighted composite value. and These represent the deviations in vibration and oil condition, respectively, with meanings consistent with the steps described above.

[0118] This represents the cooperative deviation, which is the absolute value of the difference between two deviation values. When the vibration and the oil deviation change in opposite directions, the cooperative deviation is introduced to... Automatically increases to prevent false negatives caused by the cancellation of positive and negative deviations. When the directions of change are the same, .

[0119] In this embodiment, the introduction of a cooperative deviation allows the overall state assessment value to automatically increase when there are contradictory deviations between the two signal sources, thereby improving the system's sensitivity to complex and abnormal scenarios. This mechanism avoids the false negative problem caused by the mutual cancellation of positive and negative deviations in simple weighted summation.

[0120] In this embodiment, the preset threshold is determined based on the distribution of comprehensive status evaluation values ​​of normal operation and fault status in historical data. The threshold that optimizes detection accuracy and false alarm rate can be selected using ROC curve analysis. When the comprehensive status evaluation value exceeds this threshold, the system triggers an abnormal alarm, reminding maintenance personnel to inspect or maintain the reducer.

[0121] The technical solutions of the above embodiments, through the synergistic effect of the above steps, can effectively solve the problems existing in the prior art of reducer condition detection methods.

[0122] First, by employing order tracking and equal-angle resampling techniques, the non-stationary time-domain vibration signal is converted into a stationary angular-domain signal. Based on this, order spectrum features are extracted, effectively eliminating frequency ambiguity issues under variable-speed operating conditions. Furthermore, by establishing a zonal benchmark based on operating parameters, the vibration detection standard can adapt to different operating conditions, reducing the false alarm rate caused by changes in operating conditions.

[0123] Secondly, a wear type identification method based on the ratio and gradient change sequence of ferromagnetic to non-ferromagnetic particle content was developed, enabling refined differentiation of different wear mechanisms. Furthermore, phased modeling of oil characteristic benchmarks was implemented in conjunction with equipment operation stages, allowing oil detection standards to adapt to dynamic changes throughout the equipment's lifecycle.

[0124] Furthermore, through an adaptive weight allocation mechanism during operation, the fusion weights of vibration and oil signals are dynamically adjusted according to the equipment's break-in, stabilization, or degradation stage, enabling the comprehensive evaluation to fully utilize the most indicative signal sources at each stage. By introducing data confidence levels to correct the weights, the interference of low-quality data on the evaluation results is reduced. The collaborative deviation detection mechanism enhances the ability to capture contradictory abnormal signs.

[0125] In summary, the above embodiments of this application achieve comprehensive and accurate detection of the operating status of the reducer, improve the sensitivity and accuracy of fault detection, reduce the false alarm rate and missed alarm rate, and provide reliable technical support for the preventive maintenance and health management of the reducer.

[0126] Specifically, such as Figure 10 As shown, step S56, which involves determining whether the directions of change of the vibration state deviation value and the oil state deviation value are opposite, includes: S561. Obtain the vibration state deviation value sequence and oil state deviation value sequence obtained from the state assessment calculation of the most recent preset number of times; S562. Perform linear regression on the vibration state deviation value sequence and the oil state deviation value sequence respectively, and calculate the vibration deviation slope and the oil deviation slope. If the product of the vibration deviation slope and the oil deviation slope is negative, the direction of change is determined to be opposite; otherwise, the direction of change is determined to be the same.

[0127] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0128] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting the operating status of a speed reducer based on big data, characterized in that, The method includes: Acquire real-time and historical operating data of the reducer, and obtain historical vibration and historical oil data based on the historical operating data; The order spectrum features are obtained from the historical vibration data, and the vibration feature benchmark under historical operating conditions is generated based on the statistical distribution of the order spectrum features. The oil particle concentration trend is obtained based on the historical oil data, and an oil characteristic benchmark under normal operating conditions is generated based on the oil particle concentration trend. Based on the real-time operating data, real-time vibration data and real-time oil data are obtained, and the current operating condition parameters corresponding to the real-time vibration data are obtained. Based on the current operating condition parameters, the corresponding operating condition interval is matched from the vibration characteristic benchmark to obtain the target benchmark. The rotational speed pulse signal and vibration acceleration signal in the real-time vibration data are acquired, and the real-time vibration data is subjected to order tracking and equal angle resampling to obtain the real-time angular domain vibration signal. The real-time angular domain vibration signal is subjected to fast Fourier transform to extract real-time order spectrum features, and the real-time order spectrum features are grouped by order according to the order correlation range in the target reference. Calculate the degree of deviation between the amplitude of each order within each order group and the normal fluctuation range of the corresponding order in the target benchmark; The coupling deviation of the degree of deviation between adjacent orders within each order group is calculated based on the order association range, wherein the coupling deviation is used to reflect the degree of cooperative anomaly of the deviation between adjacent orders. The deviation degree and coupling deviation of each order group are weighted and summarized according to a preset weight to obtain the vibration state deviation value, and the oil state deviation value is obtained according to the real-time oil data and oil characteristic benchmark. A comprehensive state evaluation value is generated based on the vibration state deviation value and the oil state deviation value, and it is determined whether the comprehensive state evaluation value is greater than a preset threshold. If the comprehensive status evaluation value is less than or equal to the preset threshold, the reducer is determined to be in normal operating condition. If the comprehensive status assessment value is greater than the preset threshold, the reducer is determined to be in an abnormal operating state.

2. The method for detecting the operating status of a speed reducer based on big data according to claim 1, characterized in that, The step of obtaining the order spectrum features based on the historical vibration data includes: Multiple sets of rotational speed pulse signals and vibration acceleration signals are obtained based on the historical vibration data. Instantaneous rotational speed is calculated based on the rotational speed pulse signals. The historical vibration data is then segmented based on the fluctuation rate of the instantaneous rotational speed to obtain several vibration data segments. The rotational speed fluctuation rate is obtained from the vibration data segment, and the vibration data segment with the rotational speed fluctuation rate less than the preset fluctuation threshold is determined as the steady-state vibration data segment. For each steady-state vibration data segment, order tracking is performed, and the steady-state vibration data segment is resampled at equal angles according to the rotational speed pulse signal to obtain the angular domain vibration signal; The angular domain vibration signal is subjected to a fast Fourier transform to obtain an order spectrum, and an amplitude sequence within a preset order range is extracted from the order spectrum. The coefficient of variation of the amplitude sequence of the same order is obtained from the steady-state vibration data segment. The order amplitude with the coefficient of variation less than the preset variation threshold is determined as the effective order amplitude, and the effective order amplitude is used as the order spectrum feature.

3. The method for detecting the operating status of a speed reducer based on big data according to claim 2, characterized in that, The step of generating vibration characteristic benchmarks under historical operating conditions based on the statistical distribution of the order spectrum features includes: The transmission chain parameters of the reducer are obtained, the vibration excitation source corresponding to each order is identified according to the transmission chain parameters, and the order spectrum features are divided into at least two order feature subsets according to the type of the vibration excitation source. Obtain the operating parameters corresponding to the historical operating data, wherein the operating parameters include speed and load; divide the historical operating data into multiple operating intervals according to the operating parameters; Within each of the aforementioned operating condition intervals, the mean and standard deviation of the amplitude of each order in each of the aforementioned order feature subsets are calculated respectively; Based on the mean and the standard deviation, determine the normal fluctuation range of each order amplitude within each operating condition range; Calculate the ratio of adjacent order amplitudes in each of the order feature subsets, and statistically analyze the distribution range of the ratios to determine the order correlation range of each of the order feature subsets; The normal fluctuation range of each order amplitude within each operating condition interval and the order correlation range of each order feature subset are used together as the vibration characteristic reference range.

4. The method for detecting the operating status of a speed reducer based on big data according to claim 1, characterized in that, The step of obtaining the oil particle concentration trend based on the historical oil data includes: Obtain particle count and particle composition data from the historical oil data, wherein the particle count includes the number of particles in multiple preset size ranges, and the particle composition data includes the content of ferromagnetic particles and the content of non-ferromagnetic particles; The ratio of ferromagnetic particles to non-ferromagnetic particles is calculated based on the particle composition data, and the wear type is identified based on the ratio, wherein the wear type includes abrasive wear, fatigue wear, and corrosive wear. The historical oil data is classified according to the wear type to obtain the oil data corresponding to each wear type; Sliding window sampling was performed on the oil data corresponding to each of the wear types, and the slope of the change in the number of particles in each size interval within each sampling window was calculated. Arrange the slopes of change in chronological order to obtain the particle concentration change trends for each size range under each wear type, and use these particle concentration change trends as the oil particle concentration trends.

5. The method for detecting the operating status of a speed reducer based on big data according to claim 4, characterized in that, The step of generating an oil characteristic benchmark under normal operating conditions based on the oil particle concentration trend includes: The equipment runtime corresponding to the oil particle concentration trend is obtained, and the oil particle concentration trend is divided into the break-in stage, the stabilization stage and the degradation stage based on the equipment runtime. The median and interquartile range of particle concentration for each size range are obtained based on the oil particle concentration trends corresponding to the break-in stage, stabilization stage, and degradation stage. The upper and lower limits of particle concentration for each size range are obtained based on the median and the interquartile range. The ratio change is obtained based on the particle concentration in adjacent size ranges, and the ratio change is subjected to time-series smoothing to obtain a size distribution stability index. Based on the break-in stage, stabilization stage and degradation stage, the frequency percentage of each wear type is obtained, and the wear type with a frequency percentage greater than a preset percentage threshold is determined as the dominant wear type. The upper and lower limits of the reference, the size distribution stability index, and the dominant wear type of each size range are used together as the oil characteristic reference.

6. The method for detecting the operating status of a speed reducer based on big data according to claim 5, characterized in that, The step of obtaining the oil state deviation value based on the real-time oil data and oil characteristic benchmark includes: Obtain the current equipment runtime corresponding to the real-time oil data, and match the upper limit, lower limit, size distribution stability index and dominant wear type of the corresponding stage from the oil feature benchmark; Obtain the real-time particle count and real-time particle composition data from the real-time oil data, and obtain the amount by which the real-time particle concentration in each size range exceeds the corresponding upper limit of the oil characteristic benchmark based on the real-time particle count and real-time particle composition data. The real-time content ratio of ferromagnetic particles to non-ferromagnetic particles is obtained based on the real-time particle composition data, and the current wear type is identified based on the real-time content ratio. Determine whether the current wear type is consistent with the dominant wear type of the corresponding stage. If they are inconsistent, set the wear type deviation to a preset deviation value; if they are consistent, the wear type deviation is zero. The ratio change of adjacent size intervals in the real-time particle concentration is obtained, and the ratio change is compared with the size distribution stability index in the oil characteristic benchmark to obtain the size distribution deviation. After normalizing the excess amount, wear type deviation amount, and size distribution deviation amount of each size range, the values ​​are weighted and summarized according to preset weights to obtain the oil state deviation value.

7. The method for detecting the operating status of a speed reducer based on big data according to claim 6, characterized in that, The step of obtaining the real-time content ratio of ferromagnetic particles to non-ferromagnetic particles based on the real-time particle composition data, and identifying the current wear type based on the real-time content ratio, includes: Based on the real-time particle composition data, the ferromagnetic proportion of the number of ferromagnetic particles in each of the multiple preset size ranges is calculated as the proportion of ferromagnetic particles to the total number of particles in that size range. Based on the ferromagnetic ratio of each size interval, the change in the ferromagnetic ratio between adjacent size intervals is calculated to obtain a gradient change sequence. The ferromagnetic proportion and gradient change sequence of each size range are compared with preset wear type determination conditions to determine the current wear type. The wear type determination conditions include the ferromagnetic proportion range corresponding to abrasive wear, the gradient change range corresponding to fatigue wear, and the non-ferromagnetic particle proportion threshold corresponding to corrosion wear.

8. The method for detecting the operating status of a speed reducer based on big data according to claim 1, characterized in that, The step of generating a comprehensive condition assessment value based on the vibration state deviation value and the oil state deviation value includes: The current device runtime corresponding to the real-time operating data is obtained, and the current operating stage of the reducer is determined based on the current device runtime, wherein the operating stage includes the break-in stage, the stabilization stage and the degradation stage; The vibration reference weight and the oil reference weight are determined according to the operating stage, wherein, during the break-in stage, the oil reference weight is greater than the vibration reference weight; during the stabilization stage, the vibration reference weight is greater than the oil reference weight; and during the degradation stage, the vibration reference weight and the oil reference weight tend to be balanced. Obtain the vibration data confidence level of the real-time vibration data and the oil data confidence level of the real-time oil data; Based on the confidence levels of the vibration data and the oil data, the vibration reference weights and the oil reference weights are corrected respectively to obtain the final vibration weights and the final oil weights. The vibration state deviation value is weighted by the final vibration weight, the oil state deviation value is weighted by the final oil weight, and the weighted vibration state deviation value is added to the weighted oil state deviation value to obtain a weighted comprehensive value. Determine whether the directions of change of the vibration state deviation value and the oil state deviation value are opposite; If the directions of change are the same, then the weighted composite value is determined as the comprehensive state evaluation value; If the directions of change are opposite, the absolute value of the difference between the vibration state deviation value and the oil state deviation value is calculated as the cooperative deviation amount, and the weighted comprehensive value is added to the cooperative deviation amount to obtain the comprehensive state evaluation value.

9. The method for detecting the operating status of a speed reducer based on big data according to claim 8, characterized in that, The step of determining whether the direction of change of the vibration state deviation value and the oil state deviation value is opposite includes: Obtain the vibration state deviation sequence and oil state deviation sequence obtained from the most recent preset number of state assessment calculations; Linear regression was performed on the vibration state deviation value sequence and the oil state deviation value sequence respectively to calculate the vibration deviation slope and the oil deviation slope; If the product of the vibration deviation slope and the oil deviation slope is negative, the direction of change is determined to be opposite; otherwise, the direction of change is determined to be the same.

Images