A method and system for predicting failure of a speed reducer based on multi-sensor fusion

Abstract

The application provides a kind of based on multisensor fusion's speed reducer fault prediction method and system, belongs to the field of fault prediction, including S1, obtains the pre-processing of speed reducer operating data, obtains standardization to be detected data, and carries out hierarchical feature extraction;S2, constructs working condition reference library, compares multilevel feature with working condition reference library, determines the current working condition category of speed reducer, and carries out dynamic correction, obtains the deviation value of each level feature;S3, the deviation value of each level feature is evaluated by multisensor cooperation, and the comprehensive fault risk score is obtained;S4, according to the comprehensive fault risk score, dynamic fault diagnosis is carried out, and the fault type of speed reducer is obtained;S5, the fault type of speed reducer and operating data are stored in fault library, and based on maintenance strategy self-adaption, maintenance strategy suggestion is generated.The application realizes the accurate prediction and classification of speed reducer fault.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault prediction, in particular to a reducer fault prediction method and system based on multi-sensor fusion. BACKGROUND

[0002] Reducers are widely used in industrial production, energy, transportation and other fields, and are the core transmission components in rotating machinery. Their main function is to reduce the speed of power sources such as motors and increase the output torque, achieving power characteristic transformation. Reducers usually contain one or more gear transmission systems, which can achieve speed reduction and torque increase by changing the diameter ratio of the gears.

[0003] As a key link in mechanical systems, the operating state of the reducer directly affects the safety and efficiency of the entire production line. In existing reducer fault detection methods, most use fixed thresholds, which cannot adapt to normal parameter changes of the reducer under different loads, and the reducer is prone to false alarms in normal operating conditions such as start-stop process and load fluctuations. This results in poor operating condition adaptability of fault detection. Moreover, existing reducer fault detection usually collects single data, such as measuring the vibration signals of gears, bearings and other components for fault recognition, detecting abnormal hot spots using infrared thermography or temperature sensors, and determining the device state by detecting wear particles in the oil. However, the early stage signals of reducer faults are weak, and single data signals are difficult to reliably determine. In addition, different types of faults (such as minor tooth surface scratches and early bearing damage) exhibit similar behavior on a single sensor, leading to incorrect fault classification.

[0004] Therefore, finding a method that can distinguish between the device state of the reducer under complex operating conditions and improve the accuracy of reducer fault identification is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0005] The present application provides a reducer fault prediction method and system based on multi-sensor fusion to solve the single data fault classification defects in the prior art, achieve accurate prediction and classification of reducer faults, and effectively solve the adaptability problem of reducer fault detection under different operating conditions.

[0006] The present application provides a reducer fault prediction method based on multi-sensor fusion, comprising the following steps:

[0007] S1, obtaining the operating data of the reducer through multiple sensors, preprocessing the operating data of the reducer to obtain standardized detection data, and performing hierarchical feature extraction on the standardized detection data to obtain multi-level features;

[0008] S2, a working condition benchmark library is constructed, the multi-level characteristics are compared with the working condition benchmark library respectively, a current working condition category of the speed reducer is determined, and the multi-level characteristics are dynamically corrected according to the current working condition category of the speed reducer, so that a deviation value of each level characteristic is obtained;

[0009] S3, the deviation values of the level characteristics are evaluated by multi-sensor cooperation, and a comprehensive fault risk score is obtained;

[0010] S4, dynamic fault diagnosis is performed according to the comprehensive fault risk score, and a fault type of the speed reducer is obtained;

[0011] S5, the fault type and operation data of the speed reducer are stored in a fault library, and a maintenance strategy suggestion is generated based on adaptive maintenance strategy.

[0012] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, the multi-level characteristics include mechanical level characteristics, thermodynamic level characteristics and operation condition level characteristics, wherein,

[0013] The mechanical level characteristics include vibration energy characteristics, vibration envelope characteristics and frequency spectrum characteristics;

[0014] The thermodynamic level characteristics include temperature absolute value, temperature change rate and temperature gradient;

[0015] The operation condition level characteristics include load characteristics, rotating speed characteristics and oil characteristics.

[0016] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, step S2 specifically includes:

[0017] The working condition of the current operation state of the speed reducer is identified based on the multi-level characteristics, and a current working condition category is obtained; wherein the current operation state includes a start transition working condition, a light load steady state working condition, a medium load steady state working condition, a heavy load working condition, an overload working condition and a braking working condition;

[0018] The historical operation data of the speed reducer is obtained, a working condition benchmark library is constructed according to the historical operation data of the speed reducer, the working condition benchmark library is queried based on the current working condition category, the normal reference range of each level characteristic of the current working condition category is obtained, and the multi-level characteristic values are compared with the normal reference range of each characteristic, so that a working condition comparison result is obtained;

[0019] The working condition comparison result is calibrated by dynamic correction, and the deviation value of each level characteristic is obtained.

[0020] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, the dynamic correction includes cross-influence correction, historical trend correction and seasonal environment correction, wherein,

[0021] The cross-influence correction is used to adjust the correlation between the hierarchical features, to obtain cross-corrected feature values;

[0022] The historical trend correction is used to correct the cross-corrected feature values according to a baseline formed according to long-term operation data of the speed reducer, to obtain trend-corrected feature values;

[0023] The seasonal environment correction is used to correct the trend-corrected feature values according to the current environmental temperature conditions, to obtain final hierarchical feature deviation values.

[0024] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, the cross-influence correction specifically includes:

[0025] A multi-sensor feature correlation matrix of the speed reducer is constructed based on the multi-level features, and the correlation between the hierarchical features under the current working condition category is identified according to the multi-sensor feature correlation matrix of the speed reducer;

[0026] The correlation strength between each hierarchical feature and other hierarchical features is calculated based on the correlation, and the correction amount of each hierarchical feature is determined;

[0027] Each hierarchical feature is corrected according to the correction amount of each hierarchical feature, to obtain cross-corrected feature values.

[0028] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, the historical trend correction specifically includes:

[0029] The historical operation data of the speed reducer is obtained, and an aging baseline of the hierarchical features changing with operation time is extracted from the historical operation data of the speed reducer, and the aging baseline is fitted to obtain trend curve parameters;

[0030] The aging features in the multi-level feature values under the current working condition category are removed according to the trend curve parameters, to obtain trend-corrected feature values.

[0031] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, step S3 specifically includes:

[0032] The deviation values of the hierarchical features in the gear meshing period of the speed reducer are synchronously detected through a sliding time window, the correlation degree of the hierarchical feature deviation values in time and space is analyzed, and it is determined whether the hierarchical features are derived from the same speed reducer fault source;

[0033] Different mappings are adopted according to the number of speed reducer fault sources, to obtain real-time multi-sensor collaborative scores;

[0034] According to historical operation data of the speed reducer, a speed reducer fault mode library is constructed, the hierarchical features are matched with the speed reducer fault mode library, and a speed reducer historical fault mode matching score is generated;

[0035] The real-time multi-sensor collaborative score is fused with the speed reducer historical fault mode matching score to obtain a comprehensive fault risk score.

[0036] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, step S4 specifically comprises:

[0037] A risk threshold is set, when the comprehensive risk score exceeds the risk threshold, a candidate fault set is formed based on the hierarchical features of the speed reducer and the speed reducer fault mode library, and an initial confidence is assigned to each fault type in the candidate fault set;

[0038] The operation data of the next time window of the speed reducer is continuously collected and analyzed to obtain a current speed reducer fault development trend;

[0039] Based on the current speed reducer fault development trend and the development trend of each fault type in the candidate fault set, the confidence of the current speed reducer fault type is adjusted;

[0040] According to the confidence of the current speed reducer fault type and the initial confidence of each fault type in the candidate fault set, the fault type and severity of the current working condition category are determined.

[0041] According to the speed reducer fault prediction method based on multi-sensor fusion provided by the application, the maintenance strategy self-adaptation specifically comprises:

[0042] According to the fault type and severity of the current working condition category, the remaining time for the speed reducer fault to develop to a critical state is evaluated, which is recorded as the remaining use time of the speed reducer;

[0043] Based on the remaining use time of the speed reducer, a differentiated maintenance suggestion is generated for the speed reducer fault;

[0044] The state repair of the speed reducer after each maintenance operation is recorded, and the maintenance effect of the speed reducer fault is evaluated, the evaluation standard of the fault type is automatically adjusted based on the maintenance effect of the speed reducer fault, and the maintenance suggestion of the speed reducer fault is dynamically updated according to the service life of the speed reducer and the environmental changes.

[0045] The application also provides a speed reducer fault prediction system based on multi-sensor fusion, which realizes the speed reducer fault prediction method as described above, comprising:

[0046] a data processing module, configured to acquire operation data of the speed reducer by the multiple sensors, pre-process the operation data of the speed reducer to obtain standardized to-be-detected data, and perform hierarchical feature extraction on the standardized to-be-detected data to obtain multiple hierarchical features;

[0047] a dynamic correction module, configured to construct a working condition reference library, compare the multiple hierarchical features with the working condition reference library respectively, determine a current working condition category of the speed reducer, and perform dynamic correction on the multiple hierarchical features respectively according to the current working condition category of the speed reducer to obtain deviation values of the hierarchical features;

[0048] a risk assessment module, configured to perform multiple-sensor collaborative amplification on the deviation values of the hierarchical features to obtain a comprehensive fault risk score;

[0049] a fault classification module, configured to perform dynamic fault diagnosis according to the comprehensive fault risk score to obtain a fault type of the speed reducer;

[0050] a strategy generation module, configured to store the fault type of the speed reducer and the operation data into a fault library, and generate a maintenance strategy suggestion based on adaptive maintenance strategies.

[0051] The speed reducer fault prediction method and system based on multiple-sensor fusion provided by the application realize accurate prediction and classification of speed reducer faults by dynamically correcting and collaboratively evaluating multi-dimensional features of mechanical, thermodynamic and operation condition levels, effectively solve the adaptability problem of speed reducer fault detection under different working conditions by using a working condition reference library and dynamic correction, and significantly improve the sensitivity and accuracy of early fault identification of the speed reducer by performing spatio-temporal correlation analysis on multiple-sensor information. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0053] Figure 1 is a flowchart of the speed reducer fault prediction method based on multiple-sensor fusion provided by the application. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0055] like Figure 1 As shown, this invention provides a method for predicting gearbox faults based on multi-sensor fusion, comprising the following steps:

[0056] S1. Obtain the operating data of the reducer through multiple sensors, preprocess the operating data of the reducer to obtain standardized data to be detected, and perform hierarchical feature extraction on the standardized data to be detected to obtain multi-level features.

[0057] The multi-sensor system includes vibration sensors, temperature sensors, torque sensors, and acoustic emission sensors. Vibration sensors are placed in the reducer bearing housing and gearbox housing to capture gear meshing vibrations and abnormal bearing rolling. Temperature sensors are installed in the lubricating oil lines or bearing cavities to measure temperature in real time and reflect the thermal state. Torque sensors are installed on the input or output shaft to identify large fluctuations in load conditions (such as starting shocks, overloads, etc.). An acoustic emission transmitter is also installed to collect the impact sound signals of microcracks in the reducer.

[0058] Understandably, while gear bearing failures and gear failures in a reducer may share some similar vibration characteristics, they differ significantly in temperature characteristics and load response. Furthermore, early fault signals are typically weak and may be masked by single-level feature extraction. Therefore, this invention, by extracting hierarchical features from the reducer's operating data, can distinguish between fault types that exhibit similar behaviors but have different mechanisms, thus amplifying weak fault signals. For example, minute gear cracks may not be obvious in vibration characteristics, but by combining acoustic emission characteristics and load fluctuation characteristics for synergistic analysis, detection sensitivity can be significantly improved.

[0059] In one embodiment of the present invention, the preprocessing includes digital filtering, digital normalization and signal segmentation. Bandpass filtering is used to remove high-frequency noise and low-frequency interference signals, the data from different sensors are adjusted to a uniform range, and the data segments are divided according to the sampling frequency and the working cycle of the reducer to improve data processing efficiency.

[0060] Specifically, the multi-level features include mechanical level features, thermodynamic level features, and operating condition level features, wherein,

[0061] The mechanical hierarchical features include vibration energy features, vibration envelope features, and spectral features;

[0062] The thermodynamic level features include absolute temperature, temperature change rate and temperature gradient.

[0063] The operating condition level features include load feature, rotating speed feature and oil feature.

[0064] It can be understood that the mechanical level is mainly aimed at processing vibration data and acoustic emission data, the mechanical level features can directly reflect the physical quantity of the mechanical state of the speed reducer, the thermodynamic level is mainly aimed at processing temperature data, the thermodynamic level features can reflect the index of energy conversion and heat balance state of the speed reducer, and the operating condition level is mainly aimed at processing torque data and operating parameters, and the operating condition level features embody the index of the overall working state of the speed reducer.

[0065] The speed reducer operating data is divided into the mechanical level, the thermodynamic level and the operating condition level for feature extraction in the application, and multi-dimensional perception of the speed reducer state is realized.

[0066] In an embodiment of the application, the mechanical level feature extraction includes vibration energy feature extraction, vibration envelope feature extraction and spectrum feature extraction, wherein the vibration energy feature extraction specifically includes: calculating time domain statistical parameters, including root mean square value, peak value, peak-to-peak value, peak factor, pulse factor and margin factor, dividing the vibration signal into multiple frequency bands (low frequency 0-500Hz, medium frequency 500-2000Hz and high frequency 2000-5000Hz), calculating the energy proportion of each frequency band, for the acoustic emission signal, calculating the acoustic energy cumulative value and counting rate for early fatigue crack detection; the vibration envelope feature extraction specifically includes: performing envelope analysis on the vibration signal of the speed reducer by using Hilbert transform, extracting modulation signal features, calculating the amplitude at the bearing inner / outer ring characteristic frequency, cage characteristic frequency and gear meshing characteristic frequency in the envelope spectrum, and extracting envelope modulation depth and modulation index for evaluating the rolling element state of the speed reducer bearing; the spectrum feature extraction specifically includes: performing fast Fourier transform on the vibration signal of the speed reducer, calculating the gear meshing frequency and its harmonic amplitude, extracting the gear meshing frequency sideband energy and center frequency energy ratio of the speed reducer, calculating the spectrum kurtosis and spectrum entropy value, quantifying the spectrum shape feature, and performing cepstrum analysis on the spectrum, and extracting the periodic fault of the speed reducer gear.

[0067] In an embodiment of the present application, the extraction of the thermodynamic level features includes temperature absolute value feature extraction, temperature change rate feature extraction and temperature gradient feature extraction, wherein the temperature absolute value feature extraction specifically includes: extracting the average temperature and the maximum temperature of each measuring point (input bearing, output bearing, gear box body, lubricating oil) of the speed reducer, calculating the difference between the temperature of each measuring point and the ambient temperature, and extracting the duration ratio of the temperature exceeding the preset threshold; the temperature change rate feature extraction specifically includes: calculating the temperature change rate of each measuring point of the speed reducer in a short time window (10 minutes), extracting the maximum value, the average value and the standard deviation of the temperature change rate, identifying the abnormal temperature rising trend, and calculating the fluctuation coefficient of the temperature change rate; the temperature gradient feature extraction specifically includes: calculating the temperature difference between the input / output bearings of the speed reducer and the temperature difference between different positions of the gear box body, constructing a temperature gradient vector, and extracting the temperature gradient change trend.

[0068] In an embodiment of the present application, the operating condition level feature extraction includes load feature extraction, rotating speed feature extraction and oil feature extraction, wherein the load feature extraction specifically includes: calculating the average torque, the peak torque and the torque fluctuation coefficient of the speed reducer, extracting the load change rate and the load cycle characteristics (frequency, amplitude), and calculating the input / output power ratio of the speed reducer; the rotating speed feature extraction specifically includes: extracting the average rotating speed, the rotating speed fluctuation amplitude and the rotating speed change mode of the speed reducer, calculating the acceleration and deceleration values, identifying the start-stop characteristics of the speed reducer, and extracting the response relationship between the rotating speed change and the vibration and temperature change; the oil feature extraction specifically includes: extracting the oil temperature, pressure and flow parameters according to the oil sensor data, calculating the oil viscosity change index, and extracting the oil contamination index to evaluate the lubrication quality.

[0069] S2, constructing a working condition reference library, comparing the multi-level features with the working condition reference library respectively, determining the current working condition category of the speed reducer, and dynamically correcting the multi-level features according to the current working condition category of the speed reducer to obtain the deviation value of each level feature.

[0070] Further, step S2 specifically includes:

[0071] Based on the multi-level features, the current operating state of the speed reducer is identified to obtain the current working condition category; wherein the current operating state includes a start-up transition working condition, a light-load steady-state working condition, a medium-load steady-state working condition, a heavy-load working condition, an overload working condition and a braking working condition;

[0072] The historical operation data of the speed reducer is acquired, the working condition reference library is constructed according to the historical operation data of the speed reducer, the normal reference range of each level feature of the current working condition category is acquired by querying the working condition reference library based on the current working condition category, and the multi-level feature values are compared with the normal reference range of each feature to obtain the working condition comparison result;

[0073] The deviation degree value of each level characteristic is obtained by calibrating the comparison result of the working conditions by using dynamic correction.

[0074] The calculation formula of the deviation degree value is:

[0075]

[0076] The calculation formula of the deviation degree value is: The initial deviation degree value of the i-th level characteristic is represented by i, The measured value of the i-th level characteristic is represented by i, The reference mean value of the i-th level characteristic under the current working condition category is represented by i, The reference standard deviation of the i-th level characteristic under the current working condition category is represented by i.

[0077] It can be understood that the reference mean value and the reference standard deviation can be determined according to historical operation data statistics.

[0078] In an embodiment of the present application, the speed characteristic and the load characteristic in the operation condition level characteristic are used for working condition category determination, that is, the running stage of the speed reducer is identified by analyzing the slope characteristics of the speed time curve. When it is detected that the speed continuously rises rapidly and the torque fluctuation is large, it is determined that it is "start-up transition working condition". When it is detected that the speed continuously decreases and is accompanied by energy feedback characteristics (such as the input shaft torque appearing negative or direction reversal for a short time, current direction change or voltage parameter anomaly, etc.), it is determined that it is "braking working condition". For the stage with relatively stable speed, it is classified as a steady-state working condition, and further subdivided based on the load rate.

[0079] For the subdivision of the steady-state working condition, the load rate is obtained by calculating the ratio of the current average torque to the rated torque of the speed reducer. When the load rate is not more than 30%, the current working condition is determined as "light load steady-state working condition"; when the load rate is between 30% and 70%, it is determined as "medium load steady-state working condition"; when the load rate is between 70% and 100%, it is determined as "heavy load working condition"; when the load rate is more than 100%, it is determined as "overload working condition". In actual application, the working condition judgment result will also be corrected in combination with the torque fluctuation coefficient and the oil temperature and other characteristics to improve the accuracy of working condition identification. For example, for a speed reducer at a certain moment, the speed characteristic shows that the input shaft speed is stable at about 1450 rpm, the torque fluctuation coefficient is 0.15, and the average torque is 64% of the rated torque, so the working condition is determined as "medium load steady-state working condition". If the speed of the speed reducer at another moment shows a significant downward trend, and is accompanied by a short-term negative torque characteristic, it is determined as "braking working condition".

[0080] It can be understood that the specific data basis of working condition identification and the normal reference range of each level characteristic in the working condition reference library can be set according to the actual use of the speed reducer, and the present application does not make specific limitations.

[0081] A specific embodiment is used to illustrate the working condition recognition, where the transmission ratio of the speed reducer is 4.5, the rated torque is 5000 N·m, and the rated input speed is 1480 rpm:

[0082] The multi-level features of the speed reducer are extracted, including: average speed 1450 rpm, average torque 3200 N·m, torque fluctuation coefficient 0.15, bearing RMS vibration value 1.8 mm / s, gear meshing frequency amplitude 0.9 g, input bearing temperature 58°C, and ambient temperature 25°C;

[0083] Based on the speed and torque features, it is determined that the speed reducer is currently in a steady state running stage, and the load rate is calculated as 64% (3200 / 5000×100%). It is a medium load steady state working condition.

[0084] The working condition reference library is queried to obtain the normal reference range of each feature under the medium load steady state working condition: bearing RMS vibration value range [0.7, 2.1] mm / s (mean 1.4, standard deviation 0.23), gear meshing frequency amplitude range [0.4, 1.2] g (mean 0.8, standard deviation 0.13), and input bearing temperature range [45, 65] °C (mean 55, standard deviation 3.33).

[0085] The initial deviation degrees of each level feature are calculated: the bearing RMS vibration value deviation degree is 0.58, the gear meshing frequency amplitude deviation degree is 0.26, and the input bearing temperature deviation degree is 0.30.

[0086] The initial deviation degrees of each level feature are dynamically corrected: considering that the ambient temperature is 25°C, which is higher than the reference ambient temperature 20°C, the input bearing temperature deviation degree is corrected to -0.10; considering that the device has been running for 18 months, the reference mean of the bearing RMS vibration value is corrected to 1.48 mm / s, and the corresponding deviation degree is corrected to 0.46; considering that the torque fluctuation coefficient is 0.15, the gear meshing frequency amplitude deviation degree is corrected to 0.23.

[0087] The corrected deviation degrees of each level feature are output: the bearing RMS vibration value deviation degree is 0.46, the gear meshing frequency amplitude deviation degree is 0.23, and the input bearing temperature deviation degree is -0.10.

[0088] In an embodiment of the present application, the dynamic correction includes cross-influence correction, historical trend correction, and seasonal environment correction, wherein,

[0089] The cross-influence correction is used to adjust the correlation between each level feature to obtain a cross-corrected feature value.

[0090] The historical trend correction is used to correct the cross correction characteristic value according to a baseline formed according to long-term operation data of the speed reducer, to obtain a trend correction characteristic value;

[0091] The seasonal environment correction is used to correct the trend correction characteristic value according to a current environment temperature condition, to obtain a final characteristic deviation degree value of each level.

[0092] The application can dynamically adjust the evaluation standard of the characteristic deviation degree through the cross influence correction, the historical trend correction and the seasonal environment correction, and effectively eliminates the interference of the working condition change, the equipment aging and the environment temperature on the fault judgment of the speed reducer.

[0093] Further, the cross influence correction specifically includes:

[0094] Based on the multi-level characteristics, a multi-sensor characteristic correlation matrix of the speed reducer is constructed, and the correlation between the characteristics of each level under the current working condition category is identified according to the multi-sensor characteristic correlation matrix of the speed reducer; the calculation formula is:

[0095]

[0096] Among them, Pearson correlation coefficient between level characteristic i and level characteristic j, Value of level characteristic i in the kth sample, Mean value of level characteristic i, Total number of samples, Value of level characteristic j in the kth sample, Mean value of level characteristic j;

[0097] Based on the correlation, the correlation strength between each level characteristic and other level characteristics is calculated, and the correction amount of each level characteristic is determined; the calculation formula is:

[0098]

[0099] Among them, Cross influence correction amount of level characteristic i, Influence factor, Current deviation degree value of level characteristic j, Reference deviation degree value of level characteristic j;

[0100] According to the correction amount of each level characteristic, each level characteristic is corrected to obtain a cross correction characteristic value, and the calculation formula is:

[0101]

[0102] Among them, cross correction characteristic value of the hierarchical characteristic i, initial deviation value of the hierarchical characteristic i.

[0103] It can be understood that, when cross influence correction is performed when the deviation value is greater than 0.5. Wherein the influence factor It can be set according to actual conditions, and needs to be adjusted for different working condition categories. The reference deviation value is set according to the reference value under normal state.

[0104] In an embodiment of the present application, the method for constructing the characteristic correlation matrix is: for each working condition category, the Pearson correlation coefficient between different characteristics is calculated to form an n×n correlation matrix R (n is the total number of characteristics), wherein the matrix element represents the correlation coefficient between the hierarchical characteristic i and the hierarchical characteristic j, and if is greater than a preset threshold value, it is considered that the hierarchical characteristic i and the hierarchical characteristic j have significant correlation. Wherein, the preset threshold value can be set according to actual use requirements.

[0105] Further, the historical trend correction specifically includes:

[0106] The historical running data of the speed reducer is obtained, and the aging baseline of each hierarchical characteristic with running time is extracted from the historical running data of the speed reducer, and the trend curve parameters are obtained by fitting the aging baseline; the trend curve is , wherein, represents the aging trend function, t represents the running time, a represents the trend intensity coefficient, b represents the time index, and c represents the initial offset;

[0107] According to the trend curve parameters, the aging characteristics in the multi-hierarchical characteristic value under the current working condition category are removed to obtain the trend correction characteristic value, and the calculation formula is:

[0108]

[0109]

[0110] , wherein, represents the correction reference mean value of the hierarchical characteristic i at time t, represents the initial reference mean value of the hierarchical characteristic i, represents the aging trend function value of the hierarchical characteristic i, represents the trend correction characteristic value of the hierarchical characteristic i, represents the measured value of the hierarchical characteristic i, represents the reference standard deviation of the hierarchical characteristic i.

[0111] It can be understood that the trend intensity coefficient a, the time index b and the initial offset c can be fitted according to historical operation data.

[0112] In an embodiment of the present application, the extraction method of the aging baseline is: arranging the historical data during the normal operation of the speed reducer in chronological order, calculating the feature average value of each month (or other appropriate time period) to form a feature-time sequence; using an exponential or polynomial function to fit these sequences to obtain a trend curve describing the change of the feature with time.

[0113] In an embodiment of the present application, the seasonal environment correction specifically includes:

[0114] Real-time recording of the current environment temperature, and determining the baseline environment temperature in combination with the historical operation data of the speed reducer;

[0115] Analyzing the sensitivity of the environment temperature to different levels of features, and establishing a linear relationship between the level features and the environment temperature;

[0116] According to the linear relationship, the different levels of features are corrected to obtain the final deviation value of each level of feature, and the calculation formula is:

[0117]

[0118]

[0119]

[0120] Among them, sensitivity coefficient of feature i to environment temperature, deviation value of level feature i, environment temperature variable, environment temperature correction amount of level feature i, current environment temperature, baseline environment temperature, final corrected deviation value, trend correction feature value of level feature i, environment temperature correction amount.

[0121] It can be understood that the baseline environment temperature is determined according to historical operation data, and is usually 20℃. The sensitivity coefficient is set according to actual use.

[0122] S3, the deviation values of the level features are evaluated in coordination with multiple sensors to obtain a comprehensive fault risk score.

[0123] Specifically, step S3 specifically includes:

[0124] Synchronous detection of the deviation degree values of the hierarchical features in the gear meshing period of the speed reducer is performed through a sliding time window, and the correlation of the deviation degree values of the hierarchical features in time and space is analyzed to determine whether the hierarchical features originate from the same speed reducer fault source;

[0125] Different mappings are used according to the number of speed reducer fault sources to obtain real-time multi-sensor collaborative scoring;

[0126] A speed reducer fault mode library is constructed according to historical operation data of the speed reducer, similarity matching of the hierarchical features and the speed reducer fault mode library is performed, and a speed reducer historical fault mode matching score is generated;

[0127] The real-time multi-sensor collaborative score and the speed reducer historical fault mode matching score are fused to obtain a comprehensive fault risk score.

[0128] As can be understood, each fault mode in the speed reducer fault mode library contains the deviation degree value distribution, time evolution law and mutual relationship of the multi-level features at different stages of fault development. For example, for bearing faults, the mode library records typical patterns of vibration features, temperature changes, lubrication states and efficiency drops at each stage from early micro-cracks to severe damage.

[0129] The present application realizes the collaborative amplification of weak fault signals by analyzing the correlation of different hierarchical features in time and space. In the early fault stage (such as gear micro-cracks), the single sensor signal may not reach the alarm threshold, but through multi-sensor collaborative amplification, potential faults can be detected in advance, providing sufficient preparation time for preventive maintenance.

[0130] In an embodiment of the present application, when the speed reducer fault source is a single fault source, a weighted summation mapping is used. For example, bearing faults usually cause characteristic frequency vibration, temperature rise and lubrication state change, and a higher weight (such as 0.5) is given to the vibration feature, a medium weight (such as 0.3) is given to the temperature feature, and a lower weight (such as 0.2) is given to the operating condition feature. The deviation degree values of the hierarchical features are multiplied by the response weights and summed to obtain the collaborative score of the single fault source.

[0131] In an embodiment of the present application, when the speed reducer fault source is multiple fault sources, a nonlinear mapping method is used. For example, the maximum priority method is used, that is, the score of each fault source is calculated separately, and then the maximum value is taken as the main score, and a mutual influence factor is added to reflect the risk aggravation caused by the coexistence of multiple faults.

[0132] In an embodiment of the present application, the hierarchical features are matched with the reducer failure mode library in terms of similarity, and the similarity calculation method can be based on Euclidean distance, Mahalanobis distance or cosine similarity, and the present application does not make specific limitations in this regard.

[0133] In an embodiment of the present application, the real-time multi-sensor collaborative scoring and the reducer historical failure mode matching score are fused, and the fusion adopts adaptive weight fusion, that is, the weights of the two scores are dynamically adjusted according to the quality of the real-time multi-sensor data and the similarity of the reducer historical failure mode, for example, when the signal-to-noise ratio of the real-time multi-sensor data is high and the sensor is in good working condition, the weight of the real-time multi-sensor collaborative scoring is increased; when the current working condition is highly similar to the reducer failure mode library, the weight of the reducer historical failure mode matching score is increased.

[0134] S4, dynamically diagnosing a failure type of the reducer according to the comprehensive failure risk score.

[0135] Specifically, the step S4 specifically comprises:

[0136] setting a risk threshold, when the comprehensive risk score exceeds the risk threshold, forming a candidate failure set based on the hierarchical features of the reducer and the reducer failure mode library, and assigning an initial confidence to each failure type in the candidate failure set;

[0137] continuously collecting and analyzing the operation data of the next time window of the reducer to obtain a current failure development trend of the reducer;

[0138] adjusting the confidence of the current failure type of the reducer based on the current failure development trend of the reducer and the development trend of each failure type in the candidate failure set;

[0139] determining the failure type and severity of the current working condition according to the confidence of the current failure type of the reducer and the initial confidence of each failure type in the candidate failure set.

[0140] wherein the next time window can be 24 hours or one week, and the present application does not make specific limitations in this regard.

[0141] The present application adopts a dynamic failure diagnosis method of continuously collecting and analyzing the reducer data, which can dynamically evaluate the failure development trend and adjust the confidence of the failure type, thereby improving the accuracy of failure type identification.

[0142] It can be understood that the initial confidence is determined based on the matching degree of the current hierarchical feature and the typical feature in the reducer failure mode library. The present application adopts a fuzzy matching algorithm to calculate the similarity of the current feature mode and each failure mode in the reducer failure mode library, and converts the similarity into the initial confidence. The confidence value ranges from 0 to 100%, indicating the degree of certainty of the failure type judgment. For example, if the input bearing outer ring fails and the feature matching degree reaches 70%, the initial confidence of the bearing outer ring may be set to 65%; if the abnormality of the gear meshing frequency is also observed at the same time, "gear wear" is a candidate failure, but the matching degree is only 40%, and the initial confidence of the gear wear may be only 35%.

[0143] In an embodiment of the present application, the current reducer failure development trend mainly includes the change rate, change direction and change mode of the feature deviation. By calculating the first derivative (change rate) and the second derivative (change acceleration) of the feature deviation, the speed and acceleration / deceleration trend of the failure development are judged. At the same time, the evolution mode of the correlation between different hierarchical features over time is analyzed, such as the change of the time sequence correlation between mechanical vibration features and temperature features.

[0144] In an embodiment of the present application, the failure type with the highest confidence and exceeding a certain threshold in the candidate failure set is determined as the main failure type. If the confidence of multiple failure types exceeds the threshold, it is determined as a compound failure. The failure severity is divided into four levels: slight (confidence > 75%, feature deviation < 0.7, and slow growth), moderate (confidence > 80%, feature deviation 0.7-0.85, or moderate growth rate), severe (confidence > 85%, feature deviation > 0.85, or rapid growth), and critical (confidence > 90%, feature deviation close to 1, and very rapid growth).

[0145] S5, storing the failure type and operation data of the reducer into a failure library, and generating a maintenance strategy suggestion based on adaptive maintenance strategy.

[0146] The adaptive maintenance strategy specifically includes:

[0147] According to the failure type and severity of the current working condition category, the remaining time for the reducer failure to develop to a critical state is evaluated, which is recorded as the remaining use time of the reducer;

[0148] Based on the remaining use time of the reducer, a differentiated maintenance suggestion is generated for the reducer failure:

[0149] For early failure of the reducer, it is suggested to increase the monitoring frequency and make preventive planning;

[0150] For the intermediate failure of the reducer, it is suggested to shut down for maintenance in the next time window;

[0151] Suggesting immediate shutdown for repair for late failure of the speed reducer;

[0152] Recording the state repair of the speed reducer after each maintenance operation, and evaluating the failure maintenance effect of the speed reducer, automatically adjusting the evaluation standard of the failure type based on the failure maintenance effect of the speed reducer, and dynamically updating the failure maintenance suggestion of the speed reducer according to the service life of the speed reducer and environmental changes.

[0153] According to the fault type and severity, a differentiated maintenance suggestion is automatically generated, and the fault evaluation standard is dynamically adjusted based on the maintenance effect, and the balance optimization of maintenance cost and equipment reliability is realized based on the adaptive maintenance strategy.

[0154] It can be understood that the remaining time for evaluating the critical state of the failure needs to consider various factors, including: the current working condition category (the failure development rate is different under different working conditions), environmental conditions (such as temperature, humidity, etc.), the service life of the speed reducer (the failure development of old speed reducers is usually faster), and the historical maintenance situation (frequent failure of components may exist systematic problems).

[0155] The present application realizes accurate prediction and classification of speed reducer failure by dynamically correcting and co-analyzing multi-dimensional features of mechanical level, thermodynamic level and operating condition level, effectively solves the adaptability problem of speed reducer failure detection under different working conditions by adopting working condition benchmark library and dynamic correction; at the same time, through the spatio-temporal correlation analysis of multi-sensor information, the sensitivity and accuracy of early failure identification of the speed reducer are significantly improved, which provides a reliable basis for the maintenance decision of the speed reducer, reduces the probability of unplanned shutdown of the equipment, prolongs the service life of the speed reducer, and reduces the maintenance cost.

[0156] The present application also provides a speed reducer failure prediction system based on multi-sensor fusion, which realizes the speed reducer failure prediction method as described above, comprising:

[0157] A data processing module is used for acquiring the operating data of the speed reducer through multi-sensor, pre-processing the operating data of the speed reducer to obtain standardized detection data, and performing hierarchical feature extraction on the standardized detection data to obtain multi-level features.

[0158] A dynamic correction module is used for constructing a working condition benchmark library, comparing the multi-level features with the working condition benchmark library respectively to determine the current working condition category of the speed reducer, and dynamically correcting the multi-level features according to the current working condition category of the speed reducer to obtain the deviation value of each level feature.

[0159] A risk assessment module is used for multi-sensor cooperative amplification of the deviation value of each level feature to obtain a comprehensive failure risk score.

[0160] a fault classification module, configured to perform dynamic fault diagnosis according to the comprehensive fault risk score, to obtain a fault type of the speed reducer;

[0161] a strategy generation module, configured to store the fault type of the speed reducer and the operation data into a fault library, and to generate a maintenance strategy suggestion based on adaptive maintenance strategy.

[0162] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting gearbox faults based on multi-sensor fusion, characterized in that, Includes the following steps: S1. Acquire operating data of the reducer through multiple sensors, preprocess the operating data of the reducer to obtain standardized data to be tested, and perform hierarchical feature extraction on the standardized data to be tested to obtain multi-level features. The multi-level features include mechanical level features, thermodynamic level features, and operating condition level features, wherein... The mechanical hierarchical features include vibration energy features, vibration envelope features, and spectral features; The thermodynamic hierarchy features include absolute temperature value, rate of temperature change, and temperature gradient; The operating condition hierarchical characteristics include load characteristics, speed characteristics, and oil characteristics; S2. Construct a working condition benchmark library, compare the multi-level features with the working condition benchmark library respectively, determine the current working condition category of the reducer, and dynamically correct the multi-level features according to the current working condition category of the reducer to obtain the deviation value of each level feature. S3. Perform multi-sensor collaborative evaluation on the deviation values ​​of the features at each level to obtain a comprehensive fault risk score; S4. Perform dynamic fault diagnosis based on the comprehensive fault risk score to obtain the fault type of the reducer; S5. Store the fault types and operating data of the reducer in the fault database, and generate maintenance strategy suggestions based on the maintenance strategy adaptively.

2. The method for predicting gearbox faults based on multi-sensor fusion according to claim 1, characterized in that, Step S2 specifically includes: Based on the aforementioned multi-level features, the current operating state of the reducer is identified to obtain the current operating condition category; wherein the current operating state includes start-up transition condition, light load steady-state condition, medium load steady-state condition, heavy load condition, overload condition, and braking condition; The historical operating data of the reducer is obtained, and a working condition benchmark library is constructed based on the historical operating data of the reducer. The working condition benchmark library is queried based on the current working condition category to obtain the normal reference range of each level of feature of the current working condition category. The multi-level feature values ​​are compared with the normal reference range of each feature to obtain the working condition comparison result. Dynamic correction is used to calibrate the working condition comparison results to obtain the deviation values ​​of each level of features.

3. The method for predicting gearbox faults based on multi-sensor fusion according to claim 2, characterized in that, The dynamic correction includes cross-influence correction, historical trend correction, and seasonal environmental correction, among which, The cross-influence correction is used to adjust the correlation between features at each level to obtain cross-corrected feature values; The historical trend correction is used to correct the cross correction feature value based on the baseline formed by the long-term operation data of the reducer, so as to obtain the trend correction feature value; The seasonal environmental correction is used to adjust the trend correction feature value according to the current environmental temperature conditions, so as to obtain the final feature deviation value of each level.

4. The method for predicting gearbox faults based on multi-sensor fusion according to claim 3, characterized in that, The cross-influence correction specifically includes: Based on the multi-level features, a multi-sensor feature association matrix for the speed reducer is constructed, and the correlation between features at each level under the current working condition category is identified according to the multi-sensor feature association matrix for the speed reducer. Based on the aforementioned correlation, the correlation strength between each hierarchical feature and other hierarchical features is calculated, and the correction amount for each hierarchical feature is determined. The features at each level are corrected according to the correction amount of each level feature to obtain cross-corrected feature values.

5. The method for predicting gearbox faults based on multi-sensor fusion according to claim 4, characterized in that, The historical trend correction specifically includes: Acquire historical operating data of the reducer, and extract aging baselines of each level of features as a function of operating time from the historical operating data of the reducer. Fit the aging baselines to obtain trend curve parameters. Based on the trend curve parameters, aging features are removed from the multi-level feature values ​​under the current operating condition category to obtain trend correction feature values.

6. The method for predicting gearbox faults based on multi-sensor fusion according to claim 2, characterized in that, Step S3 specifically includes: By using a sliding time window, the deviation values ​​of each level of features within the gear meshing cycle of the reducer are simultaneously detected. The correlation between the deviation values ​​of each level of features in time and space is analyzed to determine whether each level of features originates from the same reducer fault source. Different mappings are used based on the number of fault sources in the reducer to obtain a real-time multi-sensor collaborative score; A fault mode library for the speed reducer is constructed based on the historical operating data of the speed reducer. The features at each level are matched with the fault mode library for similarity to generate a historical fault mode matching score for the speed reducer. The real-time multi-sensor collaborative score is fused with the historical fault mode matching score of the reducer to obtain a comprehensive fault risk score.

7. The method for predicting gearbox faults based on multi-sensor fusion according to claim 6, characterized in that, Step S4 specifically includes: A risk threshold is set. When the comprehensive risk score exceeds the risk threshold, a candidate fault set is formed based on the characteristics of each level of the reducer and the reducer fault mode library, and an initial confidence level is assigned to each fault type in the candidate fault set. By continuously collecting and analyzing the operating data of the reducer in the next time window, the current trend of reducer failure can be obtained; Based on the current fault development trend of the reducer and the development trend of each fault type in the candidate fault set, adjust the confidence level of the current fault type of the reducer. The fault type and severity of the current operating condition category are determined based on the confidence level of the current gearbox fault type and the initial confidence level of each fault type in the candidate fault set.

8. The method for predicting gearbox faults based on multi-sensor fusion according to claim 1, characterized in that, The adaptive maintenance strategy specifically includes: The remaining time for the gearbox to develop into a critical state is assessed based on the fault type and severity of the current operating condition category, and recorded as the remaining service life of the gearbox. Based on the remaining service life of the speed reducer, differentiated maintenance suggestions are generated for speed reducer failures; Record the condition and repair status of the reducer after each maintenance operation, evaluate the effectiveness of reducer fault maintenance, automatically adjust the evaluation criteria for fault types based on the effectiveness of reducer fault maintenance, and dynamically update reducer fault maintenance recommendations according to the service life of the reducer and environmental changes.

9. A speed reducer fault prediction system based on multi-sensor fusion, characterized in that, Implementing the gearbox fault prediction method as described in any one of claims 1-8, comprising: The data processing module is used to acquire the operating data of the reducer through multiple sensors, preprocess the operating data of the reducer to obtain standardized data to be detected, and perform hierarchical feature extraction on the standardized data to be detected to obtain multi-level features. The multi-level features include mechanical level features, thermodynamic level features, and operating condition level features, wherein... The mechanical hierarchical features include vibration energy features, vibration envelope features, and spectral features; The thermodynamic hierarchy features include absolute temperature value, rate of temperature change, and temperature gradient; The operating condition hierarchical characteristics include load characteristics, speed characteristics, and oil characteristics; The dynamic correction module is used to construct a working condition benchmark library, compare the multi-level features with the working condition benchmark library respectively, determine the current working condition category of the reducer, and dynamically correct the multi-level features according to the current working condition category of the reducer to obtain the deviation value of each level feature. The risk assessment module is used to amplify the deviation values ​​of the features at each level through multi-sensor collaboration to obtain a comprehensive fault risk score; The fault classification module is used to perform dynamic fault diagnosis based on the comprehensive fault risk score to obtain the fault type of the reducer; The strategy generation module is used to store the fault types and operating data of the reducer in the fault database, and generate maintenance strategy suggestions based on the maintenance strategy adaptively.

Images