Disclosure of Invention
The embodiment of the application provides a data processing method and a related device for a gearbox.
A method of data processing for a gear box comprising:
acquiring first working condition data and state data, wherein the first working condition data comprise time sequence data of a power input side measuring point and time sequence data of a load output side measuring point, and the state data comprise time sequence data of a gear box body side measuring point;
calculating a predicted value of a target measuring point according to the first working condition data and the state data;
when the difference value between the actual value of the target measuring point and the predicted value of the target measuring point is larger than an alarm threshold value, sending an early warning signal and acquiring second working condition data and first historical data in a preset time period before a fault early warning moment, wherein the actual value of the target measuring point is a part of the first working condition data, the second working condition data is time sequence data of a power input side and a load output side, the first historical data is time domain data of operation parameters of bearings, shafting and groups of meshing gears of a gear box, and the alarm threshold value is a preset value;
determining a target fault mode according to the second working condition data and the first historical data;
acquiring a first fault sample matrix of the target fault mode;
determining a maintenance time point of the target fault mode according to the first fault sample matrix;
and outputting prompt information, wherein the prompt information comprises the information of the target failure mode and the information of the maintenance time point.
Optionally, calculating a predicted value of the target measurement point according to the first operating condition data and the state data includes:
removing abnormal values from the first working condition data to obtain third working condition data;
performing a transformation between a time domain and a frequency domain on the status data to obtain first frequency domain data of the status data;
respectively calculating frequency band factors of a bearing, a shafting and a meshing gear of the gearbox according to the first frequency domain data;
normalizing the third working condition data to obtain fourth working condition data;
performing dimensionality reduction on the fourth working condition data and the frequency band factor to obtain a dimensionality reduction feature vector;
establishing a high-dimensional feature space through the dimension-reducing feature vector, wherein the high-dimensional feature space is a matrix of time sequence data comprising state information;
classifying the time sequence data of the high-dimensional feature space through a clustering model to obtain different types of time sequence data, wherein each type of time sequence data represents one operating state of the gearbox;
analyzing the time segments of the time sequence data under each classification through a regression model, and determining a first mathematical relation between a target measuring point and all measuring points except the target measuring point;
and calculating to obtain a predicted value of the target measuring point according to the first mathematical relationship and the time sequence data of all measuring points except the target measuring point.
Optionally, after the predicted value of the target measurement point is obtained through calculation by using the mathematical relationship and time series data of all measurement points except the target measurement point, the method further includes:
if the clustering models and the regression models are multiple, evaluating the clustering models and the regression models to select an optimal clustering model and an optimal regression model;
screening target predicted values corresponding to the optimal clustering model and the optimal regression model from the predicted values;
when the difference value between the actual value of the target measuring point and the predicted value of the target measuring point is larger than an alarm threshold value, sending an early warning signal and acquiring second working condition data and first historical data in a preset time period before the fault early warning moment, wherein the method comprises the following steps:
and when the difference value between the actual value of the target measuring point and the target predicted value is larger than an alarm threshold value, sending an early warning signal and acquiring second working condition data and first historical data in a preset time period before the fault early warning moment.
Optionally, determining a target failure mode according to the second operating condition data and the first historical data includes:
performing transformation between a time domain and a frequency domain on the first historical data to obtain second frequency domain data;
calculating a first change rate of the operating parameters of each bearing, each shafting and each group of meshing gears and a first statistical characteristic value of the operating parameters of each bearing, each shafting and each group of meshing gears based on the second frequency domain data;
classifying first input data through a clustering model to obtain early warning sample data, wherein the first input data comprise the second working condition data, the first change rate and the first statistical characteristic value;
establishing an early warning sample matrix through the early warning sample data, wherein a row vector of the early warning sample matrix comprises the first statistical characteristic value and the first change rate;
calculating a state feature vector of the early warning sample matrix;
calculating a plurality of similarity values between the state feature vector and a plurality of candidate feature vectors, each candidate feature vector characterizing a failure mode;
and determining a fault mode corresponding to the candidate feature vector with the maximum similarity value as a target fault mode, wherein the maximum similarity value is the maximum value in the similarity values.
Optionally, before the first historical data is transformed between the time domain and the frequency domain to obtain second frequency domain data, the method further includes:
acquiring fifth working condition data and second historical data representing historical faults, wherein the fifth working condition data comprises time sequence data of power input side operation parameters and time sequence data of load output side operation parameters, and the second historical data is second time domain data representing multiple faults of the gearbox;
removing abnormal values from the fifth working condition data to obtain sixth working condition data;
converting the second historical data into second frequency domain data;
calculating a second change rate of the operating parameters of the bearings, the shafting and the sets of meshing gears and a second statistical characteristic value of the operating parameters of the bearings, the shafting and the sets of meshing gears based on the second frequency domain data;
performing fault mode classification on second input data through a clustering model to obtain fault sample data of each fault set, wherein the second input data comprise the sixth working condition data, the second change rate and the second statistical characteristic value;
establishing a plurality of second fault sample matrixes of a plurality of fault modes through fault sample data of each fault set, wherein the row vector of each fault sample matrix comprises a second statistical characteristic value and a second change rate corresponding to each fault mode;
a plurality of candidate feature vectors of the second fault sample matrix is calculated.
Optionally, determining a maintenance time point of the target failure mode according to the first failure sample matrix includes:
normalizing the first fault sample matrix to obtain a third fault sample matrix, wherein the first fault sample matrix is a matrix of the early warning sample matrix and comprises time sequence data of the target fault mode;
classifying the third fault sample matrix through a clustering model to obtain classified sample data;
analyzing the classified sample data through a regression model, and determining a second mathematical relationship between the target measuring point and all measuring points except the target measuring point;
calculating to obtain a predicted value of the target measuring point according to the second mathematical relationship and sample data of all measuring points except the target measuring point;
and determining a time point corresponding to the minimum value in the predicted values which is greater than or equal to an arrival-at-time value which is a preset value for distinguishing the state of the gearbox as a maintenance time point.
Optionally, converting the second historical data into second frequency domain data includes:
converting the second historical data into second frequency domain data by fast Fourier transform;
or the like, or, alternatively,
converting the second historical data into second frequency domain data by discrete Fourier transform.
A data processing apparatus comprising:
the device comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit is used for acquiring first working condition data and state data, the first working condition data comprises time sequence data of a power input side measuring point and time sequence data of a load output side measuring point, and the state data comprises time sequence data of a gear box body side measuring point;
the calculating unit is used for calculating a predicted value of a target measuring point according to the first working condition data and the state data;
the processing unit is used for sending out an early warning signal and acquiring second working condition data and first historical data in a preset time period before a fault early warning moment when the difference value between the actual value of the target measuring point and the predicted value of the target measuring point is larger than an alarm threshold value, wherein the actual value of the target measuring point is a part of the first working condition data, the second working condition data is time sequence data of a power input side and a load output side, the first historical data is time domain data of operating parameters of bearings, shafting and groups of meshing gears of a gearbox, and the alarm threshold value is a preset value;
the determining unit is used for determining a target fault mode according to the second working condition data and the first historical data;
the obtaining unit is further configured to obtain a first fault sample matrix of the target fault mode;
the determining unit is further configured to determine a maintenance time point of the target failure mode according to the first failure sample matrix;
and the output unit is used for outputting prompt information, and the prompt information comprises the information of the target failure mode and the information of the maintenance time point.
A data processing apparatus comprising:
the system comprises a central processing unit, a memory and an input/output interface;
the memory is a transient memory or a persistent memory;
the central processor is configured to communicate with the memory and execute the instruction operations in the memory to perform the aforementioned methods.
A computer readable storage medium comprising instructions which, when executed on a computer, cause the computer to perform the aforementioned method.
According to the technical scheme, the embodiment of the application has the following advantages:
and when the difference value between the actual value of the target measuring point and the predicted value is greater than an alarm threshold value, sending an early warning signal and acquiring second working condition data and first historical data in a preset time period before the fault early warning moment. And then determining a target fault mode according to the second working condition data and the first historical data, acquiring a first fault sample matrix of the target fault mode, determining a maintenance time point of the target fault mode, and outputting prompt information. When the seedling ends are in fault, the early warning can be carried out, an early warning signal is sent out, a possible target fault mode and a corresponding maintenance time point are prompted, so that a reasonable maintenance scheme can be made, and great convenience is brought to users.
Detailed Description
The embodiment of the application provides a data processing method and a related device for a gearbox.
The working condition of the gear transmission is complex, the process that the bearing and the gear are gradually degraded from early abnormity to functional failure is adopted, and the failure early warning and diagnosis in the prior art has serious time lag and coping passivity, so that inconvenience is brought to users. The data processing method and the related device for the gearbox can help a user to make a reasonable maintenance scheme and improve production efficiency.
Referring to fig. 1, a method for processing data of a transmission according to an embodiment of the present application is described as follows, where the method for processing data of a transmission according to an embodiment of the present application includes:
101. acquiring first working condition data and state data;
the first working condition data and the state data are obtained from a monitoring information system, a distributed control system, a rotary machine monitoring management system and the like through sensors. A measurement point is understood to be a physical quantity. The first working condition data comprise time sequence data of a power input side measuring point and time sequence data of a load output side measuring point, and the state data comprise time sequence data of a gear box body side measuring point.
102. Calculating a predicted value of a target measuring point according to the first working condition data and the state data;
and after the first working condition data and the state data are obtained, calculating a predicted value of the target measuring point. Specifically, the first working condition data and the state data are preprocessed, and then a feature space is established. And performing clustering analysis and regression analysis on the feature space to predict the predicted value of the target measuring point.
103. When the difference value between the actual value of the target measuring point and the predicted value of the target measuring point is larger than an alarm threshold value, sending an early warning signal and acquiring second working condition data and first historical data in a preset time period before the fault early warning moment;
after the predicted value of the target measuring point is obtained, the predicted value is compared with the actual value of the measuring point obtained by the sensor. If the difference value between the first working condition data and the second working condition data is larger than the preset alarm threshold value, the deviation is too large, and the gear transmission deviates from the normal working state, an early warning signal is sent out to prompt a user, and the second working condition data and the first historical data in the preset time period before the fault early warning moment are obtained for subsequent operation. The actual value of the target measuring point is a part of the first working condition data, the second working condition data is time sequence data of a power input side and a load output side, and the first historical data is time domain data including operation parameters of bearings, shafting and groups of meshing gears of the gear box. The alarm threshold value is a preset value and can be preset according to experience and requirements.
104. Determining a target fault mode according to the second working condition data and the first historical data;
and after the second working condition data and the first historical data are obtained, determining a target fault mode. Specifically, the second working condition data and the first historical data are preprocessed, a first change rate of the operating parameters of the bearings, the shafting and the sets of meshing gears and a first statistical characteristic value of the operating parameters of the bearings, the shafting and the sets of meshing gears are obtained through calculation, an early warning sample matrix is established according to the first change rate and the first statistical characteristic value, and a state characteristic vector of the early warning sample matrix is calculated. And finally, comparing the similarity of the state characteristic vector and the candidate characteristic vector, wherein the fault mode corresponding to the maximum similarity is the target fault mode.
105. Acquiring a first fault sample matrix of the target fault mode;
and acquiring a first fault sample matrix corresponding to the target fault mode after the target fault mode is obtained, wherein the first fault sample matrix is a matrix comprising time sequence data of the target fault mode, and the target fault mode is the fault mode with the highest probability in the possible fault modes of the gear transmission.
106. Determining a maintenance time point of the target fault mode according to the first fault sample matrix;
and determining a maintenance time point of the target fault mode according to the first fault sample matrix. Specifically, the first failure sample matrix is subjected to normalization processing, then clustering analysis is performed to classify the first failure sample matrix, then regression analysis is performed, and a time point corresponding to the minimum value in the predicted values which are greater than or equal to the achievement value is determined as a maintenance time point.
107. Outputting prompt information;
and outputting prompt information after the maintenance time point is obtained. The prompt information includes information of the target failure mode and information of the maintenance time point, for example, the prompt information may be "warning: and when a possible fault occurs, the target fault mode is bearing loosening, and the maintenance time point is xx year xx month xx sun xx.
In the embodiment of the application, first working condition data and state data are obtained to calculate a predicted value of a target measuring point, when a difference value between an actual value and the predicted value of the target measuring point is larger than an alarm threshold value, an early warning signal is sent out, and second working condition data and first historical data in a preset time period before a fault early warning moment are obtained. And then determining a target fault mode according to the second working condition data and the first historical data, acquiring a first fault sample matrix of the target fault mode, determining a maintenance time point of the target fault mode, and outputting prompt information. When the seedling ends are in fault, the early warning can be carried out, an early warning signal is sent out, a possible target fault mode and a corresponding maintenance time point are prompted, so that a reasonable maintenance scheme can be made, and great convenience is brought to users.
Referring to fig. 2, another embodiment of a method for processing data of a transmission according to the embodiment of the present application includes:
201. acquiring first working condition data and state data;
the first working condition data and the state data are obtained from a monitoring information system, a distributed control system, a rotary machine monitoring management system and the like through sensors. A measurement point is understood to be a physical quantity. The first working condition data comprise time sequence data of a power input side measuring point and time sequence data of a load output side measuring point, and the state data comprise time sequence data of a gear box body side measuring point. Parameters of the power input side are motor current, power and the like, parameters of the gear box body side are input shaft horizontal vibration, vertical vibration and axial vibration, output shaft horizontal vibration, vertical vibration and axial vibration and the like, and parameters of the load output side are water pump, fan medium flow, density, lift and the like.
202. Removing abnormal values from the first working condition data to obtain third working condition data;
and removing abnormal values in the first working condition data to obtain third working condition data. Wherein, the abnormal value refers to an individual value in the sample, and the value of the abnormal value is obviously deviated from the rest observed values of the sample. The accuracy of the subsequent model prediction can be ensured by filtering the abnormal value in the first working condition data.
203. Performing a transformation between a time domain and a frequency domain on the status data to obtain first frequency domain data of the status data;
the acquired state data are time domain data, and in order to facilitate subsequent calculation, the state data are transformed into corresponding first frequency domain data through fast Fourier transform, so that the transformation from a time domain to a frequency domain is completed.
It will be appreciated that the state data may also be transformed into the corresponding first frequency domain data by a mathematical method such as discrete fourier transform that transforms the time domain data into the frequency domain state. The details are not limited herein.
204. Respectively calculating frequency band factors of a bearing, a shafting and a meshing gear of the gearbox according to the first frequency domain data;
and calculating the frequency band factors of a bearing, a shafting and a meshing gear of the gearbox according to the following formula by using the first frequency domain data.
Specifically, the frequency band factor representing the state of the bearing of the gear transmission can be calculated according to the inner ring passing frequency, the outer ring passing frequency, the ball rotation frequency and the bandwidth of 1-5 per mill of the retainer passing frequency of the bearings of the input shaft and the output shaft of the gear transmission respectively, and the specific formula is as follows:
the frequency band factor representing the centering and balance state of a gear box shafting can be calculated according to the bandwidth of 1% -3% of the rotating frequency of the input shaft and the output shaft of the gear box, and the specific formula is as follows:
the frequency band factor representing the gear engagement state of the gear transmission can be calculated according to the bandwidth of 3% -5% of the gear engagement frequency of the input shaft and the output shaft of the gear transmission, and the specific formula is as follows:
it is understood that other equivalent variations of the above formula may also be used to calculate the band factor, and are not limited herein.
If the gearbox has multiple bearings or countershafts, the frequency band factors of the bearings, shafting and sets of meshing gears are calculated respectively according to the same method.
205. Normalizing the third working condition data to obtain fourth working condition data;
for convenience of processing, the third working condition data is normalized to obtain fourth working condition data. The normalization is a simplified calculation method, and may specifically be linear function normalization or zero-mean normalization, which is not limited herein.
206. Performing dimensionality reduction on the fourth working condition data and the frequency band factor to obtain a dimensionality reduction feature vector;
and screening the fourth working condition data and the frequency band factors by adopting a dimension reduction method to obtain a dimension reduction characteristic vector. Wherein, the dimension reduction means that a certain mapping method is adopted to map the data points in the original high-dimensional space to the low-dimensional space. The visual benefit is that the dimensionality is reduced, calculation and visualization are convenient, and the deeper significance of the visual sense is effective information extraction synthesis and useless information abandonment.
Specifically, the dimension reduction method may be a principal component analysis method, and the non-orthogonal data is screened out by converting the third working condition data and the frequency band factor into the feature vector through orthogonal transformation. The dimension reduction method may also be a T-distribution random neighborhood embedding method, and the like, and is not limited herein.
207. Establishing a high-dimensional feature space through the dimension-reducing feature vector;
and establishing a high-dimensional feature space by using the dimension-reduced feature vector. The high-dimensional feature space is a matrix of time sequence data including state information, and the state information is related information in a normal state under different working conditions.
208. Classifying the time sequence data of the high-dimensional feature space through a clustering model to obtain different types of time sequence data;
in order to process the established high-dimensional feature space, a clustering model can be further adopted to classify the high-dimensional feature space. And classifying the state to which the time sequence data belongs to obtain the time sequence data of each classification, wherein the time sequence data of each classification represents an operation state of the gear transmission. The clustering model is an analysis method for researching classification problems, and on the basis of similarity, time series data in one classification have more similarity than time series data not in the same classification.
Specifically, the clustering model may be a fuzzy C-means clustering model, a density-based noise application space clustering model, a K-means clustering model, or the like, and is not limited herein.
The K-means clustering model is described here as an example. The K-means clustering model divides the data set into a plurality of different categories according to the data characteristics existing in the data, so that the data in the categories are relatively similar, and the data similarity between the categories is relatively low. Selecting n initialized class centers, calculating the distance from each data in the first input data to the class center, and setting the class of each data as the class of the class center closest to the class center, such as the class of bearing inner ring wear or the class of ball wear. And replacing the previous category center with the mean value of each category to become a new category center, and repeating the steps until a preset termination condition is reached.
209. Analyzing the time segments of the time sequence data under each classification through a regression model, and determining a first mathematical relation between a target measuring point and all measuring points except the target measuring point;
and analyzing the time segments of the time sequence data under each classification by adopting a regression model to determine a first mathematical relationship between the target measuring point and all measuring points except the target measuring point. Among them, the regression model is a predictive modeling technique that studies the relationship between dependent variables and independent variables. And predicting the values of other measuring points which are not acquired through the acquired measuring point values by adopting a regression model so as to obtain the predicted value of the measuring point to be predicted.
Specifically, the regression model may be an approximate nearest neighbor search model, a long-term memory network model, or a hopfeldt neural network model, and is not limited herein.
210. Calculating to obtain a predicted value of the target measuring point according to the first mathematical relationship and time sequence data of all measuring points except the target measuring point;
and substituting the time sequence data of all the measuring points except the target measuring point into a mathematical expression representing the relationship between the target measuring point and all the measuring points except the target measuring point to calculate the predicted value of the target measuring point. And predicting the values of other measuring points which are not acquired through the acquired time sequence data of the measuring points to obtain the predicted value of the target measuring point. For example, 10 time series data are collected from 10 sensors, and after the processing of the foregoing steps, if a predicted value of a first measuring point is to be obtained, the time series data of the other 9 measuring points are analyzed, that is, the predicted value of the first measuring point is taken as a dependent variable, and the time series data of the other 9 measuring points is taken as an independent variable.
211. If the clustering models and the regression models are multiple, evaluating the clustering models and the regression models to select an optimal clustering model and an optimal regression model;
when there are a plurality of clustering models and regression models, there are a plurality of predicted values. And evaluating the model by adopting a Receiver Operating Characteristic (ROC) method to select an optimal clustering model and an optimal regression model. For example, additional time series data is prepared in advance, prediction is performed using a clustering model classification and a regression model, and the result is compared with the result that has been derived by the clustering model and the regression model. And evaluating by an ROC method, wherein the corresponding clustering model and regression model with the minimum error of the two results are the optimal clustering model and the optimal regression model.
212. Screening target predicted values corresponding to the optimal clustering model and the optimal regression model from the predicted values;
and screening the target predicted value from the plurality of predicted values, namely the value corresponding to the optimal clustering model and the optimal regression model, and comparing the values in the subsequent process.
213. When the difference value between the actual value of the target measuring point and the target predicted value is larger than an alarm threshold value, sending an early warning signal and acquiring second working condition data and first historical data in a preset time period before the fault early warning moment;
after the target predicted value is obtained, the target predicted value is compared with the actual value of the measuring point obtained by the sensor. If the difference value between the first working condition data and the second working condition data is larger than the preset alarm threshold value, the deviation is too large, and the gear transmission deviates from the normal working state, an early warning signal is sent out to prompt a user, the second working condition data and the first historical data in a preset time period before the fault early warning moment are obtained, and the subsequent operation is carried out. The actual value of the target measuring point is a part of first working condition data, the second working condition data is time sequence data of a power input side and a load output side, the first historical data is time domain data comprising operation parameters of bearings, shafting and groups of meshing gears of the gearbox, and the alarm threshold value is a preset value and can be preset according to experience and requirements.
214. Acquiring fifth working condition data and second historical data representing historical faults;
fifth operating condition data and second historical data characterizing the historical fault are obtained by the sensor. The fifth working condition data comprises time sequence data of power input side operation parameters and time sequence data of load output side operation parameters, and the second historical data is second time domain data representing multiple faults of the gear transmission. The fifth operating condition data and the second historical data are data of previously existing failure modes, which are used for subsequent comparison.
215. Removing abnormal values from the fifth working condition data to obtain sixth working condition data;
and after the fifth working condition data is obtained, in order to make the calculation more accurate, removing abnormal values from the fifth working condition data to obtain sixth working condition data.
216. Converting the second historical data into second frequency domain data;
the obtained second historical data is time domain data, and for convenience of calculation, the second historical data is transformed into corresponding second frequency domain data through mathematical transformation, so that the transformation from the time domain to the frequency domain is completed.
217. Calculating a second change rate of the operating parameters of the bearings, the shafting and the sets of meshing gears and a second statistical characteristic value of the operating parameters of the bearings, the shafting and the sets of meshing gears based on the second frequency domain data;
and after second frequency domain data are obtained, calculating second change rates of the operating parameters of the bearings, the shafting and the sets of the meshing gears and second statistical characteristic values of the operating parameters of the bearings, the shafting and the sets of the meshing gears based on the second frequency domain data.
Specifically, the second change rate and the second statistical characteristic value of six of the inner ring passing frequency of the bearing of the gearbox, the outer ring passing frequency of the bearing, the ball rotation frequency, the cage passing frequency, the rotation frequency of the shafting and the gear meshing frequency are respectively calculated based on the second frequency domain data, for example, the second change rate and the second statistical characteristic value are calculated based on the amplitude-time relation curves of the six. The second statistical characteristic value comprises a maximum value, a minimum value, a mean value, a median value, a mean square error, a kurtosis, a skewness and the like.
If the gear box has a plurality of bearings or intermediate shafts except the input shaft and the output shaft, the change rate and the statistical characteristic value of each bearing, each shafting and each group of meshed gears are respectively calculated according to the method.
218. Fault mode classification is carried out on the second input data through a clustering model to obtain fault sample data of each fault set;
after the second input data is obtained, fault mode classification can be carried out on the second input data through the clustering model, and sample data of each fault set is obtained. And the second input data comprise sixth working condition data, a second change rate and a second statistical characteristic value. The clustering model is a statistical analysis method for classification problems and consists of a plurality of modes. Cluster analysis is based on similarity, with more similarity between patterns in one cluster than between patterns not in the same cluster.
It can be understood that the clustering model may be a fuzzy C-means clustering model, a density clustering model, a K-means clustering model, or the like, and is not limited herein.
219. Establishing a plurality of second fault sample matrixes of a plurality of fault modes through fault sample data of each fault set;
and establishing a plurality of fault sample matrixes by using the sample data of each fault set. And the row vector of each fault sample matrix comprises a second statistical characteristic value and a second change rate corresponding to each fault mode.
220. Calculating a plurality of candidate eigenvectors of the second fault sample matrix;
the plurality of candidate eigenvectors of the second failure sample matrix may be calculated by a standard deviation method, a mean square deviation method, or the like. And the candidate characteristic vector is a characteristic vector of the second fault sample matrix and represents a corresponding fault mode.
And calculating the second fault sample matrix to obtain a characteristic polynomial equation, and solving all roots of the characteristic polynomial equation according to the characteristic equation, wherein the roots are characteristic values. And solving a homogeneous linear equation set for each characteristic value to obtain a characteristic vector. If the candidate eigenvectors of the category corresponding to the matrix can be represented by the geometric centers thereof, a plurality of candidate eigenvectors can be obtained by calculating the distances by methods such as a standard deviation method, a mean square error method and the like.
221. Performing transformation between a time domain and a frequency domain on the first historical data to obtain second frequency domain data;
the acquired first historical data is time domain data, and for calculation convenience, the first historical data is transformed into corresponding second frequency domain data through mathematical transformation, so that the transformation from the time domain to the frequency domain is completed.
222. Calculating a first change rate of the operating parameters of each bearing, each shafting and each group of meshing gears and a first statistical characteristic value of the operating parameters of each bearing, each shafting and each group of meshing gears based on the second frequency domain data;
and after the second frequency domain data are obtained, calculating a first change rate of the operating parameters of the bearings, the shafting and the sets of the meshing gears and a first statistical characteristic value of the operating parameters of the bearings, the shafting and the sets of the meshing gears based on the second frequency domain data.
Wherein, a first rate of change and a first statistical characteristic value can be calculated based on the amplitude versus time curve of the aforementioned operating parameter, the first statistical characteristic value including but not limited to maximum value, minimum value, mean value, median, mean square error, kurtosis, skewness, etc.
223. Classifying the first input data through a clustering model to obtain early warning sample data;
after the first input data are obtained, the first input data can be classified through the clustering model, and early warning sample data are obtained. The first input data comprises first working condition data, a first change rate and a first statistical characteristic value. For example, the working condition data, the corresponding change rate and the corresponding statistical characteristic value under the wear mode of the bearing inner ring are classified into one category.
The clustering model is a statistical analysis method for classification problems and consists of a plurality of modes. Clustering analysis is based on similarity, with more similarity between patterns in one class than between patterns not in the same class.
The K-means clustering model is described here as an example. The K-means clustering model divides the data set into a plurality of different categories according to the data characteristics existing in the data, so that the data in the categories are relatively similar, and the data similarity between the categories is relatively low. Selecting n initialized class centers, calculating the distance from each data in the first input data to the class center, and setting the class of each data as the class of the class center closest to the class center, such as the class of bearing inner ring wear or the class of ball wear. And replacing the previous category center with the mean value of each category to become a new category center, and repeating the steps until a preset termination condition is reached.
224. Establishing an early warning sample matrix through the early warning sample data;
and establishing an early warning sample matrix by using the early warning sample data, wherein the row vector of the early warning sample matrix comprises a first statistical characteristic value and a first change rate.
225. Calculating a state feature vector of the early warning sample matrix;
and analyzing and calculating the early warning sample matrix to obtain a state characteristic vector. And the state characteristic vector is a characteristic vector of the early warning sample matrix and represents the state of the variable speed gearbox in the preset time period.
And calculating the early warning sample matrix to obtain a characteristic polynomial equation, and solving all roots of the early warning sample matrix according to the characteristic equation, wherein the roots are characteristic values. And solving a homogeneous linear equation set for each characteristic value to obtain a characteristic vector.
226. Calculating a plurality of similarity values between the state feature vector and a plurality of candidate feature vectors;
in order to compare the similarity between the state feature vector and the plurality of candidate feature vectors, the similarity value between the state feature vector and the candidate feature vectors may be calculated by a mathematical method such as an euler distance method, a manhattan distance method, a cosine distance method, a pearson correlation coefficient method, and a spearman (rank) correlation coefficient method. Wherein each candidate feature vector characterizes a particular failure mode and there are multiple similarity values.
The distance between the state feature vector and the candidate feature vector, i.e., the similarity, can be obtained by the euler distance method, and a plurality of similarity values can be obtained. If the scene is a two-dimensional scene, the Euler distance method is a formula for solving the linear distance between two points on the plane.
227. Determining a fault mode corresponding to the candidate feature vector with the maximum similarity value as a target fault mode;
and selecting the value with the largest similarity value from the acquired similarity values, and determining the value as the largest similarity value. And determining a target fault mode corresponding to the state feature vector according to the maximum similarity value, wherein the target fault mode is a fault mode corresponding to the candidate feature vector with the maximum similarity value.
228. Acquiring a first fault sample matrix of the target fault mode;
and acquiring a first fault sample matrix corresponding to the target fault mode after the target fault mode is obtained, wherein the first fault sample matrix is a matrix comprising time sequence data of the target fault mode, and the target fault mode is the fault mode with the highest probability in the possible fault modes of the gear transmission.
229. Normalizing the first fault sample matrix to obtain a third fault sample matrix;
for convenience of processing, the first fault sample matrix is normalized to obtain a third fault sample matrix. The normalization is a simplified calculation method, and may specifically be linear function normalization or zero-mean normalization, which is not limited herein. The first fault sample matrix is a matrix of the early warning sample matrix including the time sequence data of the target fault mode.
230. Classifying the third fault sample matrix through a clustering model to obtain classified sample data;
and after the third fault sample matrix is obtained, classifying the third fault sample matrix to obtain classified sample data. Wherein the classification sample data comprises data classified by different life times of the gearbox. A classification clustering model is a statistical analysis method of classification problems, and based on similarity, there is more similarity between patterns in one class than between patterns not in the same class. The clustering model in the embodiment of the present application may be a K-means clustering model, a density-based noise application space clustering model, or a gaussian mixture clustering model, which is not specifically limited herein.
231. Analyzing the classified sample data through a regression model, and determining a second mathematical relationship between the target measuring point and all measuring points except the target measuring point;
and analyzing the classified sample data by adopting a regression model, determining the mathematical relationship between the target measuring point and all measuring points except the target measuring point, and providing a precondition for subsequent calculation. Among them, the regression model is a predictive modeling technique, and studies the relationship between a predicted value and sample data for predicting it. Specifically, the regression model may be an approximate nearest neighbor search model, a long-term memory network model, or a hopfeldt neural network model, and is not limited herein.
232. Calculating to obtain a predicted value of the target measuring point according to the second mathematical relationship and sample data of all measuring points except the target measuring point;
and substituting the sample data of all the measuring points except the target measuring point into a mathematical relation representing the relation between the target measuring point and all the measuring points except the target measuring point to calculate the predicted value of the target measuring point. And predicting the values of other measuring points which are not acquired through the acquired data of the measuring points to obtain the predicted value of the target measuring point. For example, 10 sample data are acquired, and after the processing of the foregoing steps, if the predicted value of the first measuring point is to be obtained, the sample data of the other 9 measuring points are analyzed, that is, the predicted value of the first measuring point is taken as a dependent variable, and the sample data of the other 9 measuring points is taken as an independent variable.
233. Determining a time point corresponding to the minimum value in the predicted values which are greater than or equal to the achievement value as a maintenance time point;
after a plurality of predicted values are calculated, the minimum value in the predicted values which are greater than or equal to the arrival value is found out, and then the time point corresponding to the minimum value is determined as the maintenance time point. The achievement value is a preset value used for distinguishing the state of the gearbox and can be set by a user according to experience.
For example, the preset inner ring passing frequency of the bearing is 2, the gear box is failed when the frequency exceeds 2, and the gear box is normally operated when the frequency does not exceed 2. And if the minimum value in the predicted values which are greater than or equal to the arrival-required value is 2.05, the time point corresponding to the numerical value of 2.05 is a maintenance time point, and the time point is the best time for maintenance.
234. And outputting prompt information.
And outputting prompt information after the maintenance time point is obtained. The prompt information includes information of the target failure mode and information of the maintenance time point, for example, the prompt information may be "warning: and when a possible fault occurs, the target fault mode is bearing loosening, and the maintenance time point is xx year xx month xx sun xx.
In this embodiment, the first operating condition data and the state data are acquired to calculate a predicted value of the target measuring point, and when a difference value between an actual value and the predicted value of the target measuring point is greater than an alarm threshold, an early warning signal is sent out and second operating condition data and first historical data in a preset time period before a fault early warning moment are acquired. And then determining a target fault mode according to the second working condition data and the first historical data, acquiring a first fault sample matrix of the target fault mode, determining a maintenance time point of the target fault mode, and outputting prompt information. When the seedling ends are in fault, the early warning can be carried out, an early warning signal is sent out, a possible target fault mode and a corresponding maintenance time point are prompted, and therefore a reasonable maintenance scheme can be worked out. In addition, through operations such as normalization and abnormal value elimination, the complexity of operation is greatly reduced, the efficiency is improved, and great convenience is brought to users.
Referring to fig. 3, a data processing apparatus for a transmission according to an embodiment of the present application will be described, where the data processing apparatus for a transmission according to an embodiment of the present application includes:
the acquiring unit 301 is configured to acquire first working condition data and state data, where the first working condition data includes time sequence data of a power input side measuring point and time sequence data of a load output side measuring point, and the state data includes time sequence data of a gear box body side measuring point;
a calculating unit 302, configured to calculate a predicted value of a target measurement point according to the first operating condition data and the state data;
the processing unit 303 is configured to send an early warning signal and obtain second operating condition data and first history data within a preset time period before a fault early warning time when a difference between an actual value of the target measurement point and a predicted value of the target measurement point is greater than an alarm threshold, where the actual value of the target measurement point is a part of the first operating condition data, the second operating condition data is time sequence data of a power input side and a load output side, the first history data is time domain data of operating parameters of each bearing, each shafting, and each group of meshing gears including a gearbox, and the alarm threshold is a preset value;
a determining unit 304, configured to determine a target failure mode according to the second operating condition data and the first historical data;
the obtaining unit 301 is further configured to obtain a first failure sample matrix of the target failure mode;
the determining unit 304 is further configured to determine a maintenance time point of the target failure mode according to the first failure sample matrix;
an output unit 305, configured to output prompt information, where the prompt information includes information of the target failure mode and information of the maintenance time point.
In this embodiment of the application, the obtaining unit 301 obtains first operating condition data and state data to enable the calculating unit 302 to calculate a predicted value of a target measuring point, and when a difference value between an actual value of the target measuring point and the predicted value is greater than an alarm threshold, the processing unit 303 sends an early warning signal and obtains second operating condition data and first historical data within a preset time period before a fault early warning time. Then, the determining unit 304 determines a target failure mode according to the second operating condition data and the first history data, the obtaining unit 301 obtains a first failure sample matrix of the target failure mode, determines a maintenance time point of the target failure mode, and the output unit 305 outputs a prompt message. When the seedling ends are in fault, the early warning can be carried out, an early warning signal is sent out, a possible target fault mode and a corresponding maintenance time point are prompted, so that a reasonable maintenance scheme can be made, and great convenience is brought to users.
The functions and processes executed by the units in the data processing apparatus of this embodiment are similar to those executed by the data processing apparatus in fig. 1 to 2, and are not described again here.
Fig. 4 is a schematic structural diagram of a data processing apparatus 400 according to an embodiment of the present disclosure, where the data processing apparatus 400 may include one or more Central Processing Units (CPUs) 401 and a memory 405, and the memory 405 stores one or more application programs or data.
Memory 405 may be volatile storage or persistent storage, among other things. The program stored in the memory 405 may include one or more modules, each of which may include a sequence of instructions operating on the data processing apparatus 400. Still further, the central processor 401 may be arranged to communicate with the memory 405, executing a series of instruction operations in the memory 405 on the data processing apparatus 400.
The data processing device 400 may also include one or more power supplies 402, one or more wired or wireless network interfaces 403, one or more input-output interfaces 404, and/or one or more operating systems, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.
The central processing unit 401 may perform the operations performed by the data processing apparatus in the embodiments shown in fig. 1 to fig. 2, which are not described herein again.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and the like.