CN113033673A - Training method and system for motor working condition abnormity detection model - Google Patents
Training method and system for motor working condition abnormity detection model Download PDFInfo
- Publication number
- CN113033673A CN113033673A CN202110338340.5A CN202110338340A CN113033673A CN 113033673 A CN113033673 A CN 113033673A CN 202110338340 A CN202110338340 A CN 202110338340A CN 113033673 A CN113033673 A CN 113033673A
- Authority
- CN
- China
- Prior art keywords
- working condition
- motor
- data
- training
- working
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/34—Testing dynamo-electric machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The invention discloses a training method and a system for a motor working condition abnormity detection model, wherein the method comprises the following steps: continuously acquiring motor operation data to obtain motor working conditions until the motor operates stably; segmenting the motor working condition after the motor operates stably to obtain a stable working condition list, wherein the stable working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions; and training to obtain working condition abnormality detection models corresponding to each combined working condition in the steady-state working condition list by adopting an abnormal point isolation method according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormality detection models corresponds to the number of the combined working conditions. According to the invention, the working condition of the motor is segmented, so that effective motor operation working conditions can be obtained, the working condition abnormity detection model is obtained by training by adopting a method for isolating abnormal points, and then the motor state monitoring result is automatically and directly obtained by utilizing the model, so that the false alarm rate and the missing alarm rate can be reduced, and the accuracy of working condition abnormity detection can be improved.
Description
Technical Field
The invention relates to the technical field of motor diagnosis, in particular to a training method and a training system for a motor working condition abnormity detection model.
Background
The motor state monitoring and fault diagnosis technology is a technology for knowing and mastering the state of a motor in the use process, determining the whole or local normality or abnormality of the motor, and finding out a fault and the cause of the fault. In the prior art, a state monitoring and diagnosing device based on an electrical characteristic analysis technology is used for measuring current and voltage signals of a motor during load operation, analyzing characteristics such as frequency spectrum, harmonic wave and electrical parameters, further detecting state changes of a bearing fault, an misalignment fault, a load fault, mechanical looseness, insulation and a series of electrical and mechanical faults, and further judging the fault of the whole transmission system.
However, in the conventional technical solution, the current state of the device is taken as a research focus, and the future development trend of the device cannot be predicted, and the health state of the device cannot be systematically managed. How to effectively train to obtain a motor working condition abnormity detection model, and further determine whether the motor is in an abnormal state according to the motor working condition abnormity detection model is an important problem to be solved for monitoring the motor state and diagnosing motor faults based on motor data.
Therefore, a training method and system for a motor working condition abnormality detection model are needed.
Disclosure of Invention
The invention aims to provide a training method and a training system for a motor working condition abnormity detection model, which are used for solving the problems in the prior art, reducing the false alarm rate and the missing report rate and improving the working condition abnormity detection accuracy.
The invention provides a training method of a motor working condition abnormity detection model, which comprises the following steps:
continuously acquiring motor operation data to obtain motor working conditions until the motor operates stably;
segmenting the motor working condition after the motor operates stably to obtain a stable working condition list, wherein the stable working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions;
and training to obtain working condition abnormality detection models corresponding to each combined working condition in the steady-state working condition list by adopting a method of isolating abnormal points according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormality detection models corresponds to the number of the combined working conditions.
The above training method for the motor working condition abnormality detection model, wherein preferably, the continuous collection of motor operation data to obtain the motor working condition until the motor operates stably specifically includes:
collecting motor operation data to obtain a working condition file;
extracting a vibration characteristic signal based on the motor vibration original signal in the working condition file to obtain a characteristic file corresponding to the working condition file;
synchronizing the working condition file and the feature file;
obtaining a stable running state of the motor according to the synchronized working condition file and the synchronized feature file;
and finishing the collection of the motor operation data after the motor operation stable state is a stable state.
The above training method for the motor working condition anomaly detection model, wherein preferably, the motor working condition is segmented to obtain a steady-state working condition list, and specifically comprises:
performing preliminary segmentation on data in the working condition file according to the stable rotating speed vector to obtain a plurality of preliminary groups;
marking the current data in each preliminary group with a steady-state current tag;
and carrying out secondary segmentation on each primary group according to the steady-state current label to obtain a final segmentation result.
As above, the method for training the motor working condition anomaly detection model, preferably, the data in the working condition file is preliminarily segmented according to the stable rotation speed vector to obtain a plurality of preliminary groups, and specifically includes:
obtaining a plurality of stable rotating speed vectors from the configuration file;
according to the stable rotating speed vector is to data in the working condition file is preliminarily segmented to obtain a plurality of preliminary groups, so that data of different rotating speeds are segmented, and data of the same rotating speed are used as the same preliminary group, wherein the number of the preliminary groups corresponds to the number of the stable rotating speed vector, and in the same preliminary group, the same rotating speed working condition corresponds to a plurality of different current working conditions.
The training method of the motor working condition abnormality detection model as described above, wherein preferably, the performing secondary segmentation on each preliminary grouping according to the steady-state current label to obtain a final segmentation result specifically includes:
in each preliminary grouping, taking the data marked as the same steady-state current label as the same fine grouping;
obtaining a working condition list according to each fine grouping in each preliminary grouping, wherein the working condition list comprises combined working conditions of different rotating speeds and different currents;
and outputting the working condition list as a CSV file comprising a rotating speed variable and a current variable, wherein the CSV file comprises working conditions corresponding to all rotating speeds, and each rotating speed corresponds to a plurality of current working conditions.
The above training method for the motor working condition abnormality detection model, wherein preferably, the training is performed by using a method of isolating abnormal points according to the segmentation result of the motor working condition to obtain the working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list, and specifically includes:
taking the working condition serial numbers corresponding to all combined working conditions in the steady-state working condition list and a sixteen-dimensional feature matrix file as training data, and inputting the training data into a preset machine learning model, wherein each row of the sixteen-dimensional feature matrix represents a sample, a first column represents time information, a second column represents rotating speed, a third column represents a stable state, a fourth column represents temperature, and the rest columns represent 12 time domain and frequency domain features;
and training the preset machine learning model by using the training data by adopting a method for isolating abnormal points to obtain a working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list.
The above training method for the motor working condition abnormality detection model, wherein preferably, the method of isolating the abnormal point is adopted, the preset machine learning model is trained by using the training data, and the working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list is obtained, which specifically includes:
constructing an isolation forest consisting of a plurality of isolation trees based on the training data;
and traversing the isolated forest by using the data to be subjected to abnormal detection, calculating an abnormal score, and judging whether the abnormal data exists in the data to be subjected to abnormal detection according to the abnormal score.
The method for training the motor working condition abnormality detection model as described above, wherein preferably, the constructing a plurality of isolation trees based on the training data specifically includes:
step A1: randomly selecting psi samples from training data, and putting the psi samples into a root node of a quarantine tree;
step A2: randomly appointing a dimension, and randomly generating a cutting point p in the current node data, wherein the cutting point p is generated between the maximum value and the minimum value of the appointed dimension in the current node data;
step A3: generating a hyperplane by using a randomly generated cutting point p, and then dividing the data space of the current node into 2 subspaces;
step A4: placing data smaller than p in the specified dimension on a left sub-tree of the current node, and placing data larger than or equal to p on a right sub-tree of the current node;
step A5: recursion steps A2 and A3 in the subtree branches continue to construct new leaf nodes until there is only one data in the leaf node and the cut can no longer be continued or the tree has reached a defined height.
The training method of the motor working condition abnormality detection model as described above, wherein preferably, the traversing the forest isolation by the data to be detected abnormally, and calculating an abnormality score to determine whether there is abnormal data in the data to be detected abnormally according to the abnormality score, specifically includes:
for each sample x, the result for each tree is calculated, and the anomaly score is calculated using the formula:
where h (x) is the height of sample x in each tree, i.e. the path length, c (Ψ) is the average of the path lengths at a given number of samples Ψ, and is used to normalize the path length h (x) of sample x.
The invention also provides a training system of the motor working condition abnormity detection model by adopting the method, which comprises the following steps:
the data acquisition module is used for continuously acquiring the motor operation data to obtain the motor working condition until the motor operates stably;
the working condition splitting module is used for splitting the working conditions of the motor after the motor runs stably to obtain a stable working condition list, wherein the stable working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions;
and the model training module is used for training to obtain working condition abnormity detection models corresponding to each combined working condition in the steady-state working condition list by adopting a method for isolating abnormal points according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormity detection models corresponds to the number of the combined working conditions.
The invention provides a training method and a system for a motor working condition abnormity detection model, which are beneficial to obtaining effective motor running working conditions by segmenting the motor working conditions, obtain the working condition abnormity detection model by training by adopting a method for isolating abnormal points, further automatically and directly obtain a motor state monitoring result by utilizing the model, are beneficial to reducing false alarm rate and missing alarm rate and improving the accuracy of working condition abnormity detection.
Drawings
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described with reference to the accompanying drawings, in which:
FIG. 1 is a flowchart of an embodiment of a training method for a motor condition anomaly detection model according to the present invention;
fig. 2 is a structural block diagram of an embodiment of a training system of a motor working condition abnormality detection model provided by the present invention.
Detailed Description
Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The description of the exemplary embodiments is merely illustrative and is in no way intended to limit the disclosure, its application, or uses. The present disclosure may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art. It should be noted that: the relative arrangement of parts and steps, the composition of materials, numerical expressions and numerical values set forth in these embodiments are to be construed as merely illustrative, and not as limitative, unless specifically stated otherwise.
As used in this disclosure, "first", "second": and the like, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that the element preceding the word covers the element listed after the word, and does not exclude the possibility that other elements are also covered. "upper", "lower", and the like are used merely to indicate relative positional relationships, and when the absolute position of the object being described is changed, the relative positional relationships may also be changed accordingly.
In the present disclosure, when a specific component is described as being located between a first component and a second component, there may or may not be intervening components between the specific component and the first component or the second component. When it is described that a specific component is connected to other components, the specific component may be directly connected to the other components without having an intervening component, or may be directly connected to the other components without having an intervening component.
All terms (including technical or scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs unless specifically defined otherwise. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail, but are intended to be part of the specification where appropriate.
As shown in fig. 1, the training method for the motor working condition abnormality detection model provided in this embodiment specifically includes the following steps in an actual execution process:
and step S1, continuously collecting motor operation data to obtain the motor working condition until the motor operates stably.
In an embodiment of the method for training a motor working condition abnormality detection model according to the present invention, the step S1 may specifically include:
and step S11, collecting motor operation data to obtain a working condition file.
Specifically, an OPC server is used for collecting motor operation data from a programmable logic controller connected to a motor, and the motor operation data form the working condition file, wherein the motor operation data comprise time data, current data and rotating speed data.
And step S12, extracting vibration characteristic signals based on the motor vibration original signals in the working condition files to obtain characteristic files corresponding to the working condition files.
Specifically, firstly, preprocessing operation is carried out on a motor vibration original signal to obtain a preprocessed signal, then the preprocessed signal and a sampling signal are input into a pre-constructed feature extraction model to obtain a vibration feature signal corresponding to the motor vibration original signal, and a feature file is composed of the vibration feature signals.
Further, the vibration feature signal output by the feature extraction model comprises: time stamp, temperature, vibration time domain characteristic index and vibration frequency domain characteristic index. The vibration time domain characteristic index is generally used for reflecting the equipment state, and is used for fault monitoring and trend prediction; the vibration frequency domain characteristic index is generally used for diagnosing fault types, reasons and positions. Specifically, the vibration time domain feature indicators output by the feature extraction model include effective values, skewness indicators, variances, margin factors, crest factors, kurtosis indicators, and pulse factors. The vibration frequency domain characteristic indexes output by the characteristic extraction model comprise a spectrum variance, a spectrum mean value and a spectrum effective value.
And step S13, synchronizing the working condition file and the feature file.
In the specific implementation, firstly, respectively calculating the average value of current data and rotating speed data in motor operation data according to a working condition timestamp in a working condition file to obtain motor working condition information, wherein the motor working condition information comprises time information, rotating speed information and current information; then, correcting the characteristic file time stamp in the characteristic file based on the time difference between each working condition time stamp and the current characteristic file time stamp in the corresponding characteristic file to obtain a corrected time stamp; then, obtaining motor working condition information closest to the correction timestamp in the feature file from the working condition file to obtain target motor working condition information; and finally, combining the characteristic files into the target motor working condition information by taking the working condition timestamps as the standard to form the CSV file comprising the corresponding rotating speed, current and temperature under each timestamp and the characteristics of 12 time domains and frequency domains.
And step S14, obtaining the stable running state of the motor according to the synchronized working condition file and the feature file.
Firstly, after sampling time reaches working condition configuration duration, obtaining steady-state results of a plurality of single variables based on a ratio of a first variance evaluation index and a second variance evaluation index, wherein the first variance evaluation index corresponds to a difference between actual data of two adjacent sampling time points, the second variance evaluation index corresponds to a difference between actual data and filtered data representing the same sampling time point, and the plurality of single variables comprise rotating speed, and/or current, and/or temperature; then, the motor running stable state is obtained according to the steady state results of the plurality of single variables.
And step S15, finishing the collection of the motor operation data after the motor operation stable state is a stable state.
And when the motor operation stable state is a stable state and the time length of the feature file corresponding to each stable rotating speed information is more than or equal to the total sampling time length/sampling times aiming at each stable rotating speed information in the motor stable rotating speed vector, finishing the acquisition of the motor operation data.
And step S2, segmenting the motor working condition after the motor runs stably to obtain a steady-state working condition list, wherein the steady-state working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions.
In order to obtain effective motor operation conditions, different operation conditions of the motor need to be effectively segmented before machine learning training is carried out. The purpose of step S2 is to adaptively segment the training data file generated during the configuration phase according to a number of stable rotation speeds in the signature file, and to generate a stable condition list file.
In an embodiment of the method for training a motor working condition abnormality detection model according to the present invention, the step S2 may specifically include:
and step S21, carrying out preliminary segmentation on the data in the working condition file according to the stable rotating speed vector to obtain a plurality of preliminary groups.
Further, in an embodiment of the method for training the motor operating condition abnormality detection model according to the present invention, the step S21 may specifically include:
and step S211, acquiring a plurality of stable rotating speed vectors from the configuration file.
Step S212, preliminarily segmenting the data in the working condition file according to the stable rotating speed vector to obtain a plurality of preliminary groups, so as to segment the data of different rotating speeds, and using the data of the same rotating speed as the same preliminary group, wherein the number of the preliminary groups corresponds to the number of the stable rotating speed vector, and in the same preliminary group, the working condition of the same rotating speed corresponds to a plurality of different current working conditions.
Step S22, marking the current data in each of the preliminary groupings with a steady state current tag.
In specific implementation, first, rounding is performed on current data in each preliminary group, specifically, for all current data in each preliminary group, the current data whose difference with a target current value is within a preset range is attributed to the same working condition type, wherein the number of the working condition types is consistent with the number of the target current values. Further, exemplarily, the preset range includes-5% to 5%, and it should be noted that the target current value and the preset range are not specifically limited in the present invention; the current data in each preliminary group is then labeled with a steady-state current label based on the rounding result of the current data.
And step S23, carrying out secondary segmentation on each primary grouping according to the steady-state current label to obtain a final segmentation result.
In an embodiment of the method for training a motor working condition abnormality detection model according to the present invention, the step S23 may specifically include:
step S231, in each of the preliminary groups, using the data marked as the same steady-state current label as the same fine group.
Step S232, obtaining a working condition list according to each fine grouping in each preliminary grouping, wherein the working condition list comprises combined working conditions of different rotating speeds and different currents.
Step S233, outputting the operating condition list as a CSV file including a rotational speed variable and a current variable, where the CSV file includes operating conditions corresponding to each rotational speed, and each rotational speed corresponds to a plurality of current operating conditions.
Further, after step S1, before step S2, the method for training the machine condition abnormality detection model further includes:
setting configuration parameters to obtain a configuration file, wherein the configuration parameters comprise model training total sampling duration/sampling times, a sixteen-dimensional characteristic matrix file, a stable rotating speed vector and training model precision, and the configuration file corresponds to specific equipment.
And step S3, training to obtain working condition abnormality detection models corresponding to each combined working condition in the steady-state working condition list by adopting a method of isolating abnormal points according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormality detection models corresponds to the number of the combined working conditions.
The purpose of step S3 is to create an anomaly detection model for the corresponding operating conditions based on the different steady-state operating conditions.
Further, in an embodiment of the method for training the motor operating condition abnormality detection model according to the present invention, the step S3 may specifically include:
step S31, using the working condition serial numbers and the sixteen-dimensional feature matrix files corresponding to the combined working conditions in the steady-state working condition list as training data, and inputting the training data into a preset machine learning model, wherein each row of the sixteen-dimensional feature matrix represents a sample, a first column represents time information, a second column represents rotating speed, a third column represents a steady state, a fourth column represents temperature, and the remaining columns represent 12 time domain and frequency domain features.
And step S32, training the preset machine learning model by using the training data by adopting an abnormal point isolation method to obtain a working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list.
The existing anomaly detection method mainly provides a region of a normal sample in a feature space through description of the normal sample, and regarding a sample not in the region as an anomaly. The main disadvantage of these anomaly detection methods is that only the description of normal samples is optimized, but not the description of abnormal samples, which may cause a large number of false positives or detect only a small number of anomalies.
The abnormal detection method of the invention does not describe normal sample points, but isolates abnormal points and uses statistics to explain, and in a data space, sparsely distributed areas indicate that the probability of data occurring in the areas is low, so that the data falling in the areas can be considered abnormal. To find out which points are easily isolated, two subspaces can be generated by cutting once, assuming that the data space is cut with one random hyperplane. And then continuing to cut each subspace by using a random hyperplane, and circulating until only one data point is in each subspace. It will be appreciated that clusters of very high density can be cut many times before cutting ceases, but that points of very low density tend to stop in a subspace early on.
Further, in an embodiment of the method for training the motor operating condition abnormality detection model according to the present invention, the step S32 may specifically include:
and S321, constructing an isolation forest consisting of a plurality of isolation trees based on the training data.
Further, in an embodiment of the method for training the motor working condition abnormality detection model according to the present invention, the step S321 may specifically include:
step A1: Ψ samples were randomly selected from the training data to be placed into the root node of a quarantine tree.
Step A2: randomly appointing a dimension, and randomly generating a cutting point p in the current node data, wherein the cutting point p is generated between the maximum value and the minimum value of the appointed dimension in the current node data.
Step A3: and generating a hyperplane by using the randomly generated cutting point p, and then dividing the data space of the current node into 2 subspaces.
Step A4: and placing data smaller than p in the specified dimension in the left sub-tree of the current node, and placing data larger than or equal to p in the right sub-tree of the current node.
Step A5: recursion steps A2 and A3 in the subtree branches continue to construct new leaf nodes until there is only one data in the leaf node and the cut can no longer be continued or the tree has reached a defined height.
And S322, traversing the isolated forest by using the data to be subjected to abnormal detection, calculating an abnormal score, and judging whether abnormal data exists in the data to be subjected to abnormal detection according to the abnormal score.
Specifically, for each sample x, the result for each tree is calculated, and the anomaly score is calculated using the formula:
where h (x) is the height of sample x in each tree, i.e. the path length, c (Ψ) is the average of the path lengths at a given number of samples Ψ, and is used to normalize the path length h (x) of sample x.
According to the training method of the motor working condition abnormity detection model provided by the embodiment of the invention, the working condition of the motor is segmented, so that the effective motor running working condition is favorably obtained, the working condition abnormity detection model is obtained by training by adopting a method for isolating abnormal points, and then the motor state monitoring result is automatically and directly obtained by utilizing the model, so that the false alarm rate and the missing alarm rate are favorably reduced, and the accuracy of the working condition abnormity detection is improved.
Correspondingly, as shown in fig. 2, the present invention further provides a training system for a motor working condition anomaly detection model, including:
the data acquisition module 1 is used for continuously acquiring motor operation data to obtain motor working conditions until the motor operates stably;
the working condition splitting module 2 is used for splitting the working conditions of the motor after the motor operates stably to obtain a stable working condition list, wherein the stable working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions;
and the model training module 3 is used for training to obtain working condition abnormity detection models corresponding to each combined working condition in the steady-state working condition list by adopting a method for isolating abnormal points according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormity detection models corresponds to the number of the combined working conditions.
According to the training system of the motor working condition abnormity detection model provided by the embodiment of the invention, the working condition segmentation module is used for segmenting the working condition of the motor, so that the effective motor running working condition can be favorably obtained, the model training module adopts a method for isolating abnormal points to train and obtain the working condition abnormity detection model, and further, the model is used for automatically and directly obtaining the motor state monitoring result, so that the false alarm rate and the missing report rate can be favorably reduced, and the accuracy of the working condition abnormity detection can be improved.
It should be understood that the division of the components of the training system of the motor operating condition abnormality detection model shown in fig. 2 is only a division of logical functions, and the actual implementation may be wholly or partially integrated into one physical entity or physically separated. And these components may all be implemented in software invoked by a processing element; or may be implemented entirely in hardware; and part of the components can be realized in the form of calling by the processing element in software, and part of the components can be realized in the form of hardware. For example, a certain module may be a separate processing element, or may be integrated into a certain chip of the electronic device. Other components are implemented similarly. In addition, all or part of the components can be integrated together or can be independently realized. In implementation, each step of the above method or each component above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.
Thus, various embodiments of the present disclosure have been described in detail. Some details that are well known in the art have not been described in order to avoid obscuring the concepts of the present disclosure. It will be fully apparent to those skilled in the art from the foregoing description how to practice the presently disclosed embodiments.
Although some specific embodiments of the present disclosure have been described in detail by way of example, it should be understood by those skilled in the art that the foregoing examples are for purposes of illustration only and are not intended to limit the scope of the present disclosure. It will be understood by those skilled in the art that various changes may be made in the above embodiments or equivalents may be substituted for elements thereof without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.
Claims (10)
1. A training method of a motor working condition abnormity detection model is characterized by comprising the following steps:
continuously acquiring motor operation data to obtain motor working conditions until the motor operates stably;
segmenting the motor working condition after the motor operates stably to obtain a stable working condition list, wherein the stable working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions;
and training to obtain working condition abnormality detection models corresponding to each combined working condition in the steady-state working condition list by adopting a method of isolating abnormal points according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormality detection models corresponds to the number of the combined working conditions.
2. The motor working condition abnormality detection model training method according to claim 1, wherein the motor operation data is continuously collected to obtain motor working conditions until the motor operates stably, and specifically comprises:
collecting motor operation data to obtain a working condition file;
extracting a vibration characteristic signal based on the motor vibration original signal in the working condition file to obtain a characteristic file corresponding to the working condition file;
synchronizing the working condition file and the feature file;
obtaining a stable running state of the motor according to the synchronized working condition file and the synchronized feature file;
and finishing the collection of the motor operation data after the motor operation stable state is a stable state.
3. The motor working condition abnormality detection model training method according to claim 1, wherein the motor working condition is segmented to obtain a steady-state working condition list, and specifically comprises:
performing preliminary segmentation on data in the working condition file according to the stable rotating speed vector to obtain a plurality of preliminary groups;
marking the current data in each preliminary group with a steady-state current tag;
and carrying out secondary segmentation on each primary group according to the steady-state current label to obtain a final segmentation result.
4. The motor working condition abnormality detection model training method according to claim 3, wherein the preliminary segmentation is performed on the data in the working condition file according to the stable rotation speed vector to obtain a plurality of preliminary groupings, and specifically comprises:
obtaining a plurality of stable rotating speed vectors from the configuration file;
according to the stable rotating speed vector is to data in the working condition file is preliminarily segmented to obtain a plurality of preliminary groups, so that data of different rotating speeds are segmented, and data of the same rotating speed are used as the same preliminary group, wherein the number of the preliminary groups corresponds to the number of the stable rotating speed vector, and in the same preliminary group, the same rotating speed working condition corresponds to a plurality of different current working conditions.
5. The motor working condition abnormality detection model training method according to claim 3, wherein the secondary segmentation is performed on each preliminary group according to the steady-state current label to obtain a final segmentation result, and specifically comprises:
in each preliminary grouping, taking the data marked as the same steady-state current label as the same fine grouping;
obtaining a working condition list according to each fine grouping in each preliminary grouping, wherein the working condition list comprises combined working conditions of different rotating speeds and different currents;
and outputting the working condition list as a CSV file comprising a rotating speed variable and a current variable, wherein the CSV file comprises working conditions corresponding to all rotating speeds, and each rotating speed corresponds to a plurality of current working conditions.
6. The method for training the motor working condition abnormality detection model according to claim 1, wherein the working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list is trained by adopting a method of isolating abnormal points according to the segmentation result of the motor working condition, and specifically comprises the following steps:
taking the working condition serial numbers corresponding to all combined working conditions in the steady-state working condition list and a sixteen-dimensional feature matrix file as training data, and inputting the training data into a preset machine learning model, wherein each row of the sixteen-dimensional feature matrix represents a sample, a first column represents time information, a second column represents rotating speed, a third column represents a stable state, a fourth column represents temperature, and the rest columns represent 12 time domain and frequency domain features;
and training the preset machine learning model by using the training data by adopting a method for isolating abnormal points to obtain a working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list.
7. The motor working condition abnormality detection model training method according to claim 6, wherein the method for isolating the abnormal points is adopted, the preset machine learning model is trained by using the training data, and a working condition abnormality detection model corresponding to each combined working condition in the steady-state working condition list is obtained, and the method specifically comprises the following steps:
constructing an isolation forest consisting of a plurality of isolation trees based on the training data;
and traversing the isolated forest by using the data to be subjected to abnormal detection, calculating an abnormal score, and judging whether the abnormal data exists in the data to be subjected to abnormal detection according to the abnormal score.
8. The motor working condition abnormality detection model training method according to claim 7, wherein the constructing a plurality of isolation trees based on the training data specifically includes:
step A1: randomly selecting psi samples from training data, and putting the psi samples into a root node of a quarantine tree;
step A2: randomly appointing a dimension, and randomly generating a cutting point p in the current node data, wherein the cutting point p is generated between the maximum value and the minimum value of the appointed dimension in the current node data;
step A3: generating a hyperplane by using a randomly generated cutting point p, and then dividing the data space of the current node into 2 subspaces;
step A4: placing data smaller than p in the specified dimension on a left sub-tree of the current node, and placing data larger than or equal to p on a right sub-tree of the current node;
step A5: recursion steps A2 and A3 in the subtree branches continue to construct new leaf nodes until there is only one data in the leaf node and the cut can no longer be continued or the tree has reached a defined height.
9. The motor working condition abnormality detection model training method according to claim 8, wherein the traversing the forest isolation by the data to be subjected to abnormality detection, calculating an abnormality score, and determining whether there is abnormal data in the data to be subjected to abnormality detection according to the abnormality score specifically comprises:
for each sample x, the result for each tree is calculated, and the anomaly score is calculated using the formula:
where h (x) is the height of sample x in each tree, i.e. the path length, c (Ψ) is the average of the path lengths at a given number of samples Ψ, and is used to normalize the path length h (x) of sample x.
10. A training system for motor condition anomaly detection models using the method of any one of claims 1-9, comprising:
the data acquisition module is used for continuously acquiring the motor operation data to obtain the motor working condition until the motor operates stably;
the working condition splitting module is used for splitting the working conditions of the motor after the motor runs stably to obtain a stable working condition list, wherein the stable working condition list comprises a plurality of combined working conditions of different stable rotating speed working conditions and different current working conditions;
and the model training module is used for training to obtain working condition abnormity detection models corresponding to each combined working condition in the steady-state working condition list by adopting a method for isolating abnormal points according to the segmentation result of the working conditions of the motor, wherein the number of the working condition abnormity detection models corresponds to the number of the combined working conditions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110338340.5A CN113033673B (en) | 2021-03-24 | 2021-03-24 | Training method and system for motor working condition abnormity detection model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110338340.5A CN113033673B (en) | 2021-03-24 | 2021-03-24 | Training method and system for motor working condition abnormity detection model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113033673A true CN113033673A (en) | 2021-06-25 |
| CN113033673B CN113033673B (en) | 2022-10-25 |
Family
ID=76452820
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110338340.5A Active CN113033673B (en) | 2021-03-24 | 2021-03-24 | Training method and system for motor working condition abnormity detection model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113033673B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114267178A (en) * | 2021-12-30 | 2022-04-01 | 佳都科技集团股份有限公司 | Intelligent operation maintenance method and device for station |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11113300A (en) * | 1997-10-03 | 1999-04-23 | Toshiba Corp | Induction motor controller |
| WO2007074423A2 (en) * | 2005-12-29 | 2007-07-05 | Utstarcom, Inc. | Method and system for byte slice processing data packets at a packet switch |
| CN101907685A (en) * | 2010-06-18 | 2010-12-08 | 华北电力大学 | Method for checking correctness of operating parameters of generator set capable of adapting to change of operating conditions |
| CN107292350A (en) * | 2017-08-04 | 2017-10-24 | 电子科技大学 | The method for detecting abnormality of large-scale data |
| CN107894279A (en) * | 2017-10-25 | 2018-04-10 | 北京金风科创风电设备有限公司 | Synchronous detection method and device for vibration data and working condition data of wind generating set |
| CN109765332A (en) * | 2018-12-05 | 2019-05-17 | 国网辽宁省电力有限公司信息通信分公司 | Transformer exception value real-time detection and method for diagnosing faults based on isolation forest |
| CN110135614A (en) * | 2019-03-26 | 2019-08-16 | 广东工业大学 | It is a kind of to be tripped prediction technique based on rejecting outliers and the 10kV distribution low-voltage of sampling techniques |
| CN110149258A (en) * | 2019-04-12 | 2019-08-20 | 北京航空航天大学 | A kind of automobile CAN-bus network data method for detecting abnormality based on isolated forest |
| CN110298455A (en) * | 2019-06-28 | 2019-10-01 | 西安因联信息科技有限公司 | A kind of mechanical equipment fault intelligent early-warning method based on multivariable estimation prediction |
| CN110696990A (en) * | 2019-11-11 | 2020-01-17 | 中国船舶工业系统工程研究院 | Ship generator component influence identification method and system based on data driving |
| CN111191881A (en) * | 2019-12-13 | 2020-05-22 | 大唐东北电力试验研究院有限公司 | Thermal power generating unit industrial equipment state monitoring method based on big data |
-
2021
- 2021-03-24 CN CN202110338340.5A patent/CN113033673B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11113300A (en) * | 1997-10-03 | 1999-04-23 | Toshiba Corp | Induction motor controller |
| WO2007074423A2 (en) * | 2005-12-29 | 2007-07-05 | Utstarcom, Inc. | Method and system for byte slice processing data packets at a packet switch |
| CN101907685A (en) * | 2010-06-18 | 2010-12-08 | 华北电力大学 | Method for checking correctness of operating parameters of generator set capable of adapting to change of operating conditions |
| CN107292350A (en) * | 2017-08-04 | 2017-10-24 | 电子科技大学 | The method for detecting abnormality of large-scale data |
| CN107894279A (en) * | 2017-10-25 | 2018-04-10 | 北京金风科创风电设备有限公司 | Synchronous detection method and device for vibration data and working condition data of wind generating set |
| CN109765332A (en) * | 2018-12-05 | 2019-05-17 | 国网辽宁省电力有限公司信息通信分公司 | Transformer exception value real-time detection and method for diagnosing faults based on isolation forest |
| CN110135614A (en) * | 2019-03-26 | 2019-08-16 | 广东工业大学 | It is a kind of to be tripped prediction technique based on rejecting outliers and the 10kV distribution low-voltage of sampling techniques |
| CN110149258A (en) * | 2019-04-12 | 2019-08-20 | 北京航空航天大学 | A kind of automobile CAN-bus network data method for detecting abnormality based on isolated forest |
| CN110298455A (en) * | 2019-06-28 | 2019-10-01 | 西安因联信息科技有限公司 | A kind of mechanical equipment fault intelligent early-warning method based on multivariable estimation prediction |
| CN110696990A (en) * | 2019-11-11 | 2020-01-17 | 中国船舶工业系统工程研究院 | Ship generator component influence identification method and system based on data driving |
| CN111191881A (en) * | 2019-12-13 | 2020-05-22 | 大唐东北电力试验研究院有限公司 | Thermal power generating unit industrial equipment state monitoring method based on big data |
Non-Patent Citations (3)
| Title |
|---|
| MARIO JUEZ-GIL ET AL: "Early and extremely earlymulti-label fault diagnosis in induction motors", 《ISA TRANSACTIONS》 * |
| 洪军等: "机组在线运行优化系统及实时目标工况的确定", 《电力系统自动化》 * |
| 熊伟丽等: "一种基于多工况识别的过程在线监测方法", 《控制与决策》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114267178A (en) * | 2021-12-30 | 2022-04-01 | 佳都科技集团股份有限公司 | Intelligent operation maintenance method and device for station |
| CN114267178B (en) * | 2021-12-30 | 2023-09-26 | 佳都科技集团股份有限公司 | Intelligent operation maintenance method and device for station |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113033673B (en) | 2022-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Borghesi et al. | A semisupervised autoencoder-based approach for anomaly detection in high performance computing systems | |
| CN111262722B (en) | Safety monitoring method for industrial control system network | |
| EP2854053B1 (en) | Defect prediction method and device | |
| CN111475680A (en) | Method, device, equipment and storage medium for detecting abnormal high-density subgraph | |
| AU2019275633B2 (en) | System and method of automated fault correction in a network environment | |
| US20210042585A1 (en) | Abnormality detection device, abnormality detection method and computer readable medium | |
| CN113008583B (en) | Method and device for monitoring state and automatically alarming abnormality of rotating machine | |
| CN110297207A (en) | Method for diagnosing faults, system and the electronic device of intelligent electric meter | |
| CN116450399B (en) | Fault diagnosis and root cause positioning method for micro service system | |
| CN105577440A (en) | Network fault time location method and analyzing device | |
| CN116010485B (en) | Unsupervised anomaly detection method for dynamic period time sequence | |
| Tang et al. | A wind turbine bearing fault diagnosis method based on fused depth features in time–frequency domain | |
| CN115454778A (en) | Intelligent monitoring system for abnormal time sequence indexes in large-scale cloud network environment | |
| CN113033673B (en) | Training method and system for motor working condition abnormity detection model | |
| CN111191950B (en) | Method and device for analyzing abnormal oil temperature of gearbox of wind turbine generator | |
| CN116415126A (en) | Method, device and computing equipment for anomaly detection of doctor blades of paper machine | |
| CN117009180A (en) | Log and abnormal alarm information processing method and device | |
| CN112039907A (en) | Automatic testing method and system based on Internet of things terminal evaluation platform | |
| CN113030726A (en) | Motor state monitoring method and system based on data driving | |
| CN114322446B (en) | Cooling system fault early warning method and device, cooling system and working machine | |
| CN113075547A (en) | Motor data acquisition method and system | |
| CN111638989B (en) | Fault diagnosis method, device, storage medium and equipment | |
| CN113268552A (en) | Generator equipment hidden danger early warning method based on locality sensitive hashing | |
| CN114037010A (en) | Method and device for identifying abnormal electric quantity data | |
| Chandrashekar et al. | Anomaly detection in recorded CAN log using DBSCAN and LSTM Autoencoder |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |