CN113762344A - Fault identification method of machine tool spindle and fault identification model training method and device - Google Patents
Fault identification method of machine tool spindle and fault identification model training method and device Download PDFInfo
- Publication number
- CN113762344A CN113762344A CN202110895587.7A CN202110895587A CN113762344A CN 113762344 A CN113762344 A CN 113762344A CN 202110895587 A CN202110895587 A CN 202110895587A CN 113762344 A CN113762344 A CN 113762344A
- Authority
- CN
- China
- Prior art keywords
- fault
- model
- training
- machine tool
- type
- 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.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 130
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013528 artificial neural network Methods 0.000 claims abstract description 36
- 238000012795 verification Methods 0.000 claims description 15
- 238000000354 decomposition reaction Methods 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000010200 validation analysis Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 description 14
- 238000010586 diagram Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 6
- 238000012544 monitoring process Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 238000013526 transfer learning Methods 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000001050 lubricating effect Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 239000000306 component Substances 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000005461 lubrication Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a fault identification method of a machine tool spindle and a fault identification model training method and device. The fault identification method comprises the following steps: acquiring operation data of a machine tool spindle; inputting the operation data into a first fault identification model and a second fault identification model; the method comprises the following steps that a first fault recognition model is obtained by training a neural network according to first fault sample data corresponding to a first fault type of a machine tool spindle, a second fault recognition model is obtained by training a pre-training model according to second fault sample data corresponding to a second type of fault of the machine tool spindle, and the pre-training model is constructed according to model parameters of the first fault recognition model; and carrying out fault identification on the main shaft of the machine tool by using the first fault identification model and the second fault identification model. The second fault recognition model is obtained by training on the basis of the first fault recognition model, so that the accuracy and reliability of the second fault recognition model can be improved, and the accuracy and reliability of fault recognition of the machine tool spindle are further improved.
Description
Technical Field
The invention relates to the technical field of fault identification, in particular to a fault identification method for a machine tool spindle.
Background
As one of the most important and widely used machining devices in modern machine manufacturing, the operating state of the machine tool affects the working efficiency of the whole manufacturing system. Due to frequent usage, complex mechanical structure and severe working environment, machine tool equipment, especially a spindle system which is used as a core component of a machine tool and keeps rotating at a high speed for a long time, is generally difficult to accurately position and predict various faults (such as spindle unbalance, spindle friction and the like) through a traditional physical model, and once faults occur, the machine tool equipment is very likely to cause great property loss and serious consequences.
In recent years, with the development and popularization of artificial intelligence and machine learning technologies, a plurality of students and research and development personnel try to apply a data-driven algorithm model to the field of fault prediction and diagnosis and obtain certain achievements.
However, in the actual production process of the spindle system of the machine tool, due to the particularity of high-speed rotation and the like of the spindle system of the machine tool, it is very difficult to obtain enough fault data samples for model training for all fault types, so that the fault prediction technology based on the artificial intelligence technology is difficult to achieve the expected effect in the field, and becomes an industry pain point in the field.
Disclosure of Invention
The invention aims to overcome the defect that in the prior art, a precise model for identifying the fault of a machine tool spindle cannot be obtained because enough data samples of the machine tool spindle are difficult to obtain for model training, and provides a fault identification method of the machine tool spindle, a fault identification model training method and a fault identification model training device.
The invention solves the technical problems through the following technical scheme:
in a first aspect, a method for identifying a fault of a spindle of a machine tool is provided, including:
acquiring operation data of the machine tool spindle;
inputting the operating data into a first fault identification model and a second fault identification model; the first fault recognition model is obtained by training a neural network according to first fault sample data corresponding to a first fault type of the machine tool spindle, the second fault recognition model is obtained by training a pre-training model according to second fault sample data corresponding to a second type of fault of the machine tool spindle, and the pre-training model is constructed according to model parameters of the first fault recognition model;
and carrying out fault identification on the machine tool spindle by using the first fault identification model and the second fault identification model.
Optionally, inputting the operational data into a first fault identification model and the second fault identification model, comprising:
respectively determining fault characteristics corresponding to the first fault type and fault characteristics corresponding to the second fault type;
inputting first operation data matched with the fault characteristics corresponding to the first fault type in the operation data into the first fault identification model;
and inputting second operation data matched with the fault characteristics corresponding to the second fault type in the operation data into the first fault identification model.
Optionally, the model parameters comprise weight values;
the first fault recognition model and the pre-training model have the same weight value.
In a second aspect, a training method for a fault recognition model, the fault recognition model being used for recognizing a fault of a spindle of a machine tool, the training method comprising:
training a first neural network by using first fault sample data corresponding to a first fault type of the machine tool spindle to obtain a first fault identification model for identifying the first fault type;
constructing a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model;
and training the pre-training model by using second fault sample data corresponding to a second fault type of the machine tool spindle to obtain a second fault identification model for identifying the second fault type.
Optionally, the method further comprises:
determining a fault signature corresponding to the first fault type;
performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain first fault sample data of fault characteristics corresponding to the first fault type;
and/or, determining a fault signature corresponding to the second fault type;
and performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain second fault sample data corresponding to the fault characteristics of the second fault type.
Optionally, the model parameters comprise weight values;
constructing a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model, wherein the pre-training model comprises the following steps:
and configuring the weight value of the pre-training model according to the weight value of the first fault recognition model.
Optionally, the method further comprises:
training a second neural network by using the second fault sample data to obtain a verification model of the second fault identification model, wherein the second neural network and the pre-training model have the same network structure;
performing five-fold cross validation on the second fault identification model according to the validation model;
and determining the second fault identification model passing the verification as a final second fault identification model.
In a third aspect, there is provided a fault recognition apparatus for a spindle of a machine tool, comprising:
the acquisition module is used for acquiring the operating data of the machine tool spindle;
the input module is used for inputting the operation data into a first fault identification model and a second fault identification model; the first fault recognition model is obtained by training a neural network according to first fault sample data corresponding to a first fault type of the machine tool spindle, the second fault recognition model is obtained by training a pre-training model according to second fault sample data corresponding to a second type of fault of the machine tool spindle, and the pre-training model is constructed according to model parameters of the first fault recognition model;
and the identification module is used for carrying out fault identification on the machine tool spindle by using the first fault identification model and the second fault identification model.
Optionally, the input module comprises:
the determining unit is used for respectively determining the fault characteristics corresponding to the first fault type and the second fault type;
a first input unit, configured to input first operation data, which is matched with a fault feature corresponding to the first fault type, in the operation data, into the first fault identification model;
and the second input unit is used for inputting second operation data matched with the fault characteristics corresponding to the second fault type in the operation data into the first fault identification model.
Optionally, the model parameters comprise weight values;
the first fault recognition model and the pre-training model have the same weight value.
In a fourth aspect, there is provided a training apparatus for a fault recognition model for recognizing a fault of a spindle of a machine tool, the training apparatus comprising:
the first training module is used for training a first neural network by using first fault sample data corresponding to a first fault type of the machine tool spindle to obtain a first fault identification model for identifying the first fault type;
the building module is used for building a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model;
and the second training module is used for training the pre-training model by using second fault sample data corresponding to a second fault type of the machine tool spindle to obtain a second fault identification model for identifying the second fault type.
Optionally, the method further comprises:
the first determining module is used for determining fault characteristics corresponding to a first fault type, and performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain first fault sample data of the fault characteristics corresponding to the first fault type;
and/or the second determining module is used for determining the fault characteristics corresponding to the second fault type, and performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain second fault sample data corresponding to the fault characteristics of the second fault type.
Optionally, the model parameters comprise weight values;
the building module is specifically configured to:
and configuring the weight value of the pre-training model according to the weight value of the first fault recognition model.
Optionally, the method further comprises:
the verification module is used for training a second neural network by using the second fault sample data to obtain a verification model of the second fault identification model, wherein the second neural network and the pre-training model have the same network structure; and performing five-fold cross validation on the second fault identification model according to the validation model, and determining the second fault identification model passing validation as a final second fault identification model.
The positive progress effects of the invention are as follows: the machine tool spindle is subjected to fault recognition by using the first fault recognition model and the second fault recognition model, and the second fault recognition model is obtained by fine-tuning the first fault recognition model by using fault sample data corresponding to the second fault type on the basis of the first fault recognition model, so that the learned knowledge in the first fault type is stored and accumulated, the accuracy and reliability of the second fault recognition model can be improved, and the accuracy and reliability of the fault recognition of the machine tool spindle are improved.
Drawings
FIG. 1 is a flowchart of a method for training a fault recognition model according to an exemplary embodiment of the present invention;
fig. 2a is a flowchart of a method for identifying a fault of a spindle of a machine tool according to an exemplary embodiment of the present invention;
fig. 2b is a schematic diagram of an algorithm architecture of a fault identification method for a machine tool spindle according to an exemplary embodiment of the present invention;
FIG. 3 is a block diagram of a training apparatus for fault recognition models according to an exemplary embodiment of the present invention;
fig. 4 is a block diagram illustrating a fault recognition apparatus for a spindle of a machine tool according to an exemplary embodiment of the present invention;
fig. 5 is a schematic structural diagram of an electronic device according to an exemplary embodiment of the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are not intended to limit the scope of the invention.
Fig. 1 is a flowchart of a method for training a fault recognition model, where the trained fault model is used to recognize a fault of a spindle of a machine tool, according to an exemplary embodiment of the present invention, and the method includes the following steps:
101, training a neural network by using first fault sample data corresponding to a first fault type of a machine tool spindle to obtain a first fault identification model for identifying the first fault type.
For different fault types, corresponding to different fault characteristics for fault identification, the fault characteristics for identifying various fault types are determined according to the operation data of the machine tool spindle. The operation data of the machine tool spindle can be obtained by a sensor arranged on the machine tool spindle, for example, various sensors arranged near a bearing, and the fault characteristics of various fault types for identifying the machine tool spindle are determined through empirical analysis and calculation according to characteristic parameters extracted from signals, such as vibration, temperature, rotating speed and the like, monitored by the various sensors, so that fault sample data required by training fault identification models for identifying various fault types are determined. If the number of the fault types of the machine tool spindle is n, the corresponding fault sample data set can be represented as D1, D2, …, Dn.
The operation data includes various signals, such as vibration signals, temperature signals, rotation speed signals and the like. Since the information contained in the vibration signal is extremely rich, a large amount of characteristic information can be extracted. When fault sample data of a corresponding fault type are determined according to the operation data, respectively extracting frequency domain and time domain characteristic information in the vibration signals as the fault sample data by using order analysis and wavelet packet decomposition, wherein the extracted frequency domain characteristic information mainly comprises first frequency multiplication and second frequency multiplication amplitudes of the vibration signals in three directions; the time-frequency domain characteristic information mainly comprises low-pass coefficient standard deviation and energy distribution coefficients of sub wavelets of each order after wavelet packet decomposition; time domain signals such as rotation speed signals, temperature signals and the like can be directly used as characteristic information.
In one embodiment, after the operation data is acquired, the operation data is preprocessed, for example, an abnormal value with a null signal or an abnormal value with a serious signal interference is removed, and the preprocessed operation data is used to determine the fault sample data.
After the fault characteristics of various fault types are determined, the operation data of the machine tool spindle corresponding to the various fault characteristics in the fault state and the normal operation state are respectively obtained, and the fault sample data of the corresponding fault type is determined according to the operation data. The first fault sample data is also the fault sample data corresponding to the first fault type of the main shaft of the machine tool, and the second fault sample data is also the fault sample data corresponding to the second fault type of the main shaft of the machine tool.
In one embodiment, after the fault characteristics of each fault type are determined, parameters corresponding to each fault characteristic in the operating data are labeled to identify the fault type corresponding to the parameter. The correspondence between the parameter and the tag may be a one-to-one relationship or a one-to-many relationship.
For example, assuming that the operating Data includes a parameter a, a parameter b, a parameter c, and a parameter d, and the parameter corresponding to the fault feature a is the parameter a, the Label of the parameter a is configured as a _ Label (indicating the degree of fault and abnormality), and the Data corresponding to the parameter a is represented as fault sample Data a _ Data corresponding to the fault feature a; assuming that the parameters corresponding to the fault feature B are the parameter a, the parameter c, and the parameter d, the labels of the parameter a, the parameter c, and the parameter d are configured as B _ Label (indicating the degree of the fault and the abnormality), and the Data corresponding to the parameter a, the parameter c, and the parameter d is indicated as fault sample Data a _ Data corresponding to the fault feature B. And identifying the operation Data acquired under the normal operation state of the machine tool spindle as N _ Data, and identifying the Label of the corresponding parameter as N _ Label. By analogy, the data corresponding to all fault types of the machine tool spindle can be labeled.
When the first fault identification model is trained, first fault sample data is input into the neural network, the neural network is trained, and the first fault identification model used for identifying the first fault type is obtained. The first fault sample data is generally fault sample data with a large data volume or the largest data volume in fault sample data corresponding to various fault types. Eighty percent of fault sample data are randomly extracted to directly train the model, and the rest twenty percent of samples are used as a verification set. And (4) iteratively training the model until the model is verified by using the verification set, and stopping iteration until the precision of the model is not improved any more to obtain a trained first fault recognition model. The first fault identification model is used for identifying whether the machine tool spindle has a fault of a first fault type according to the operation data of the machine tool spindle.
In one embodiment, Adam (a first order optimization algorithm) and a classification cross entropy function are used as a training optimizer and a loss function of the model, respectively, in the model training process.
The neural network may be, but is not limited to, a deep neural network, where the number of input layer neuron nodes of the deep neural network matches the number of fault features corresponding to the first fault type, for example, if the number of fault features corresponding to the first fault type is m, the number of input layer neuron nodes should also be m. The number k of hidden layers of the deep neural network and the number of neurons of each hidden layer are represented as N1, N2, … and Nk, and in order to enable the gradient of the model to be gradually reduced in the training process, the number of neurons of each layer is gradually reduced, namely N1 is more than or equal to N2 is more than or equal to N … is more than or equal to Nk. The network parameters of the neural network, including the network structure (including the number of hidden layers, the number of nodes of neurons in the input layer, etc.) and the model parameters (initial values of weighted values), can be selected from the structures and parameters with better fault identification precision and stability through a large number of experiments.
In one embodiment, the number of normal sample data (sample data of the machine tool spindle in the normal operation state) in fault sample data corresponding to each fault type is consistent with that of fault sample data (sample data of the machine tool spindle in the fault state), so that the accuracy rate of the model is prevented from being reduced due to unbalanced sample data.
And 102, constructing a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model.
Wherein the model parameters include weight values.
The weight value of the pre-training model is the same as that of the first fault recognition model, namely the pre-training model shares the weight value with the first fault recognition model, and the weight value of the first fault recognition model is used as the initial value of the weight value of the second fault recognition model. And the number of input layer neuron nodes of the pre-training model is matched with the number of fault characteristics corresponding to the second fault type.
103, training a pre-training model by using second fault sample data corresponding to the second fault type of the machine tool spindle to obtain a second fault identification model for identifying the second fault type.
And when the second fault identification model is trained, inputting second fault sample data into the pre-training model, and further training the pre-training model to obtain a second fault identification model for identifying the second fault type. The second fault sample data is generally fault sample data with a small data volume in fault sample data corresponding to various fault types, and the second fault type and the first fault type are both faults of the machine tool spindle and have similar characteristics, so that the operation data (first fault sample data) acquired when the machine tool spindle operates in the first fault type state can also partially represent the abnormal rule when the second fault type occurs, and therefore on the basis of the first fault identification model, the first fault identification model is finely adjusted by using the fault sample data corresponding to the second fault type to obtain a second fault identification model for identifying the second fault type, and compared with the mode of training the neural network by using the second fault sample data alone, the data volume of the required second fault sample data is greatly reduced. And when the second fault identification model is trained, the first fault sample data is combined, so that the operation data is more comprehensively analyzed from high latitude, potential faults are identified, the learned knowledge from other faults is reused to a prediction model of similar faults, the number of samples required by training of a single model is greatly reduced, and the problem of industrial pain points that a large amount of fault data of a machine tool spindle system is difficult to obtain is solved.
During training, eighty percent of fault sample data is randomly extracted to directly train the model, and the rest twenty percent of samples are used as a verification set. And (5) iteratively training the model until the model is verified by using the verification set, and stopping iteration until the precision of the model is not improved any more to obtain a trained second fault recognition model. The second fault identification model is used for identifying whether the machine tool spindle has a fault of a second fault type according to the operation data of the machine tool spindle. The model has strong generalization capability and can improve the stability and accuracy of fault identification.
In one embodiment, the second fault sample data is further input into a neural network, the neural network is trained, the network structure of the neural network is the same as that of the pre-trained model, the initial weight value of each neuron node is a random value, a verification model for verifying the precision of the second fault identification model is obtained, the verification model is used for verifying the precision of the second fault identification model, specifically, five-fold cross verification can be used, training and testing are repeated for five times, and the comprehensive prediction precision of each fault identification model is calculated respectively. In step 103, the verified second fault identification model is determined as a final second fault identification model.
In the embodiment of the invention, on the basis of the first fault identification model, the first fault identification model is finely adjusted by using the fault sample data corresponding to the second fault type to obtain the second fault identification model for identifying the second fault type, and compared with the method for training the neural network by using the second fault sample data alone, the data volume of the second fault sample data is greatly reduced. And the second fault recognition model training is obtained based on the training of the first fault recognition model, and the learned knowledge in similar fault types is stored and accumulated, so that the operation data can be more comprehensively analyzed from high latitude, potential faults can be recognized, the accuracy and reliability of the second fault recognition model are improved, and the accuracy and reliability of state evaluation and service life prediction of the machine tool spindle are further improved.
In this embodiment, the failure types of the machine tool spindle may include, but are not limited to, mechanical failures such as bearing wear, spindle imbalance, bearing looseness, etc., hydraulic system failures, pneumatic system failures, and lubrication system failures. The first fault type may be any one of the above fault types, and the second fault type may also be any one of the above fault types, and the first fault type and the second fault type are different. Namely, a fault identification model for identifying the bearing wear fault can be used as a pre-training model, and fault sample data corresponding to any fault type of a main shaft unbalance fault, a bearing loosening fault, a hydraulic system fault, a pneumatic system fault and a lubricating system fault is used for carrying out fine adjustment on the pre-training model to obtain a fault identification model for identifying the corresponding fault type; or taking a fault identification model for identifying the unbalance fault of the main shaft as a pre-training model, and finely adjusting the pre-training model by using fault sample data corresponding to any fault type of a bearing abrasion fault, a bearing loosening fault, a hydraulic system fault, a pneumatic system fault and a lubricating system fault to obtain a fault identification model for identifying the corresponding fault type; or taking the fault identification model for identifying the faults of the pneumatic system as a pre-training model, and performing fine adjustment on the pre-training model by using fault sample data corresponding to any fault type of bearing wear faults, spindle imbalance, bearing loosening faults, hydraulic system faults and lubricating system faults to obtain the fault identification model for identifying the corresponding fault type.
Fig. 2a is a flowchart of a method for identifying a fault of a spindle of a machine tool according to an exemplary embodiment of the present invention, where the method includes the following steps:
The operation data can be operation data of the machine tool spindle which is pushed forward within a preset time length from the current moment.
In one embodiment, when the operation data is input into the first fault identification model and the second fault identification model, the fault characteristics corresponding to the first fault type and the fault characteristics corresponding to the second fault type are respectively determined, the first operation data matched with the fault characteristics corresponding to the first fault type in the operation data is input into the first fault identification model, and the second operation data matched with the fault characteristics corresponding to the second fault type in the operation data is input into the first fault identification model.
The second fault identification model is obtained by training the fault identification model provided by any one of the above embodiments.
And step 203, identifying faults of the machine tool spindle by using the first fault identification model and the second fault identification model.
It should be noted that the first fault identification model and the second fault identification model are only exemplary descriptions, and when the machine tool spindle is actually subjected to fault identification and diagnosis, fault identification can be performed on the machine tool spindle by combining a plurality of other fault identification models, and the training processes of the other fault identification models are similar to the training process of the second fault identification model.
And diagnosing and identifying the faults of the machine tool spindle by combining a plurality of fault identification models so as to find the operation faults of the machine tool spindle in time and take measures to avoid damage and loss caused by the faults.
Fig. 2b is a schematic diagram of an algorithm architecture of a method for identifying a fault of a spindle of a machine tool according to an exemplary embodiment of the present invention, in which a similar fault data set represents a data set of first fault sample data, a similar fault feature set represents a set of fault features extracted from the first fault sample data, a target fault data set represents a data set of second fault sample data, a target fault feature set represents a set of fault features extracted from the second fault sample data, real-time monitoring data represents operating data of the spindle of the machine tool, real-time monitoring data features represent features extracted from the operating data, and internal weight sharing of a deep neural network can be seen from the diagram, so that knowledge accumulation and multiplexing across fault categories are implemented, and by storing and multiplexing knowledge across fault categories, the data amount of training samples required by a fault identification model of a specific type is reduced, the problem of difficult industry pain point of obtaining special type's main shaft fault data (training sample) is solved.
The method comprises the steps of taking historical monitoring data and fault history records (actual fault conditions) of a numerical control machine tool in a certain workshop as training samples, taking fault characteristics extracted from the historical monitoring data as input of a prediction model shown in figure 2b, inputting the actual fault conditions as a label result of the model, dividing random data to form a training set and a test set aiming at various spindle faults, using extreme difference standardization as a non-dimensionalization means, performing characteristic fusion on data with different dimensions and scales, constructing a prediction model, performing fine adjustment on target faults through transfer learning, and realizing fault prediction on a complex spindle system by using a small amount of samples. The method comprises the following specific steps:
(1) acquiring historical monitoring data and corresponding fault historical records of a numerical control machine tool in a certain workshop, wherein the monitoring data mainly comprises vibration, temperature, rotating speed signals and the like. Removing abnormal samples with empty signal values, wherein the obtained data comprise 1608 samples under normal working conditions and 331 samples (class A faults) under a spindle unbalance fault state; and 666 samples (class B fault) in the bearing wear fault condition for a total of 2605 samples.
(2) And respectively extracting frequency domain and time-frequency domain characteristics in the vibration signal by using order analysis and DB4 wavelet packet decomposition. The main mold-in features include: 1. time domain features-temperature, rotational speed; 2. frequency domain characteristics-first and second frequency multiplication amplitudes of the vibration signal in three directions; 3. after the time-frequency domain feature-wavelet packet decomposition, the low-pass energy ratio of the fourth-order sub-wavelet, the fitting coefficients of the first-fourth-order sub-wavelets and the high-pass energy ratio are calculated in three directions. On the premise that vibration, rotating speed and temperature signals are all available, 35 feature columns (2 time domain features, 6 frequency domain features and 27 time-frequency domain features) which can represent the operation state of the equipment are obtained in total and serve as feature data sets of historical records.
(3) Marking the characteristic data set according to the fault history label, and dividing the data set by adopting up-sampling to ensure that the number of normal samples is consistent with that of fault samples in the data set corresponding to each fault, thereby avoiding the reduction of the accuracy of the model caused by unbalanced samples. Obtaining a data sample set D1 of the A-type fault, wherein the total number of the samples is 662; the data sample set D2 for a class B failure, for a total of 1332 samples.
(4) Respectively constructing deep neural networks M1 and M2 for predicting A-type faults and B-type faults, and setting hyper-parameters of the deep neural networks: the number of neuron nodes in the input layer is 35, and the number of nodes in the hidden layer is respectively as follows: 32 × 16 × 8 × 2, i.e., the first and second hidden layers contain 32 neurons, the third and fourth layers contain 16 neurons, and the fifth and sixth hidden layers contain 8 and 2 neurons, respectively. Meanwhile, Adam and a classified cross entropy function are respectively used as a training optimizer and a loss function of the model (the network structure and related parameters are obtained after a large number of experiments are verified, and the deep neural network of the structure is used, so that the fault identification method for the machine tool spindle system has better fault identification precision and stability).
(5) Using five-fold cross validation, 80 percent of the data from D1 and D2 was extracted for training models M1 and M2, respectively, and the remaining 20 percent of the data was used for testing the model results, which was repeated five times to obtain the prediction accuracy of M1 and M2 before migration learning. The results are as follows:
(6) and reconstructing a transfer learning model M1 '(the model structure is the same as that of M1) for predicting the A-type faults, wherein the model M2 for predicting the B-type faults is fully trained, and the weight values of all hidden layer neurons in the M2 are assigned to the hidden layer of the M1' to inherit the related knowledge about the B-type faults, which is obtained by training the M2.
(7) The model M1' was trained twice using D1 to self-learn from a small amount of class A fault data. Adopting fold-cross validation in the learning process, 80 percent of data is extracted from D1 and used for training the model M1 ', and the remaining 20 percent of data is used for testing the model result, and after five times of repetition, the prediction accuracy of M1' after the transfer learning is obtained. The results are as follows:
taking two types of faults of spindle unbalance and spindle friction of a machine tool spindle system as an example, by collecting small sample operation data of a vibration simulation experiment platform under a normal working condition and in two types of fault (A: spindle unbalance and B: bearing wear) states, deep neural network models M1 and M2 are respectively constructed according to the A type fault data and the B type fault data and used for independently identifying the two types of faults, wherein the identification accuracy of M1 to the A type fault is 75.5%, and the identification accuracy of M2 to the B type fault is 89.9%. According to the transfer learning, a model M1 'aiming at the A-type fault is reconstructed, the weight value of the neuron of the hidden layer M1' is shared with the neuron of M2, and the secondary training is carried out by using the data of the A-type fault on the basis of M2. Experimental results show that after the method for identifying the spindle faults of the small sample machine bed based on the transfer learning is used, the accuracy rate of identifying the A-type faults can be improved to 99.4% from 75.5% originally, the expected purpose is met, and the method can be put into use.
Corresponding to the embodiments of the training method of the fault recognition model and the fault recognition method of the machine tool spindle, the invention also provides embodiments of a training device of the fault recognition model and a fault recognition device of the machine tool spindle.
Fig. 3 is a schematic block diagram of a training apparatus for a fault recognition model, which is provided in an exemplary embodiment of the present invention, and is used for recognizing a fault of a spindle of a machine tool, where the training apparatus includes:
a first training module 31, configured to train a first neural network by using first fault sample data corresponding to a first fault type of the machine tool spindle, to obtain a first fault identification model for identifying the first fault type;
a building module 32, configured to build a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model;
and a second training module 33, configured to train the pre-training model by using second fault sample data corresponding to a second fault type of the machine tool spindle, to obtain a second fault identification model for identifying the second fault type.
Optionally, the method further comprises:
the first determining module is used for determining fault characteristics corresponding to a first fault type, and performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain first fault sample data of the fault characteristics corresponding to the first fault type;
and/or the second determining module is used for determining the fault characteristics corresponding to the second fault type, and performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain second fault sample data corresponding to the fault characteristics of the second fault type.
Optionally, the model parameters comprise weight values;
the building module is specifically configured to:
and configuring the weight value of the pre-training model according to the weight value of the first fault recognition model.
Optionally, the method further comprises:
the verification module is used for training a second neural network by using the second fault sample data to obtain a verification model of the second fault identification model, wherein the second neural network and the pre-training model have the same network structure; and performing five-fold cross validation on the second fault identification model according to the validation model, and determining the second fault identification model passing validation as a final second fault identification model.
Fig. 4 is a schematic block diagram of a fault identification apparatus for a spindle of a machine tool according to an exemplary embodiment of the present invention, the apparatus including:
an obtaining module 41, configured to obtain operation data of the machine tool spindle;
an input module 42 for inputting the operational data into a first fault identification model and a second fault identification model; the first fault recognition model is obtained by training a neural network according to first fault sample data corresponding to a first fault type of the machine tool spindle, the second fault recognition model is obtained by training a pre-training model according to second fault sample data corresponding to a second type of fault of the machine tool spindle, and the pre-training model is constructed according to model parameters of the first fault recognition model;
an identification module 43, configured to perform fault identification on the machine tool spindle using the first fault identification model and the second fault identification model.
Optionally, the input module comprises:
the determining unit is used for respectively determining the fault characteristics corresponding to the first fault type and the second fault type;
a first input unit, configured to input first operation data, which is matched with a fault feature corresponding to the first fault type, in the operation data, into the first fault identification model;
and the second input unit is used for inputting second operation data matched with the fault characteristics corresponding to the second fault type in the operation data into the first fault identification model.
Optionally, the model parameters comprise weight values;
the first fault recognition model and the pre-training model have the same weight value.
For the device embodiments, since they substantially correspond to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, and 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 modules can be selected according to actual needs to achieve the purpose of the scheme of the invention. One of ordinary skill in the art can understand and implement it without inventive effort.
Fig. 5 is a schematic diagram of an electronic device according to an exemplary embodiment of the present invention, and illustrates a block diagram of an exemplary electronic device 50 suitable for implementing embodiments of the present invention. The electronic device 50 shown in fig. 5 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiment of the present invention.
As shown in fig. 5, the electronic device 50 may be embodied in the form of a general purpose computing device, which may be, for example, a server device. The components of the electronic device 50 may include, but are not limited to: the at least one processor 51, the at least one memory 52, and a bus 53 connecting the various system components (including the memory 52 and the processor 51).
The bus 53 includes a data bus, an address bus, and a control bus.
The memory 52 may include volatile memory, such as Random Access Memory (RAM)521 and/or cache memory 522, and may further include Read Only Memory (ROM) 523.
The processor 51 executes various functional applications and data processing, such as the methods provided by any of the above embodiments, by running a computer program stored in the memory 52.
The electronic device 50 may also communicate with one or more external devices 54 (e.g., a keyboard, a pointing device, etc.). Such communication may be through an input/output (I/O) interface 55. Moreover, the model-generated electronic device 50 may also communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet) via a network adapter 56. As shown, network adapter 56 communicates with the other modules of model-generated electronic device 50 over bus 53. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the model-generating electronic device 50, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.
It should be noted that although in the above detailed description several units/modules or sub-units/modules of the electronic device are mentioned, such a division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more of the units/modules described above may be embodied in one unit/module according to embodiments of the invention. Conversely, the features and functions of one unit/module described above may be further divided into embodiments by a plurality of units/modules.
Embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the method provided in any of the above embodiments.
More specific examples, among others, that the readable storage medium may employ may include, but are not limited to: a portable disk, a hard disk, random access memory, read only memory, erasable programmable read only memory, optical storage device, magnetic storage device, or any suitable combination of the foregoing.
In a possible implementation manner, the present invention can also be implemented in a form of a program product, which includes program code, and when the program product runs on a terminal device, the program code is configured to enable the terminal device to execute steps of implementing the thermal runaway warning method for a battery replacement station as described in embodiment 1.
Where program code for carrying out the invention is written in any combination of one or more programming languages, the program code may be executed entirely on the user device, partly on the user device, as a stand-alone software package, partly on the user device and partly on a remote device or entirely on the remote device.
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that this is by way of example only, and that the scope of the invention is defined by the appended claims. Various changes and modifications to these embodiments may be made by those skilled in the art without departing from the spirit and scope of the invention, and these changes and modifications are within the scope of the invention.
Claims (11)
1. A fault identification method for a machine tool spindle is characterized by comprising the following steps:
acquiring operation data of the machine tool spindle;
inputting the operating data into a first fault identification model and a second fault identification model; the first fault recognition model is obtained by training a neural network according to first fault sample data corresponding to a first fault type of the machine tool spindle, the second fault recognition model is obtained by training a pre-training model according to second fault sample data corresponding to a second type of fault of the machine tool spindle, and the pre-training model is constructed according to model parameters of the first fault recognition model;
and carrying out fault identification on the machine tool spindle by using the first fault identification model and the second fault identification model.
2. The method of identifying a fault in a spindle of a machine tool according to claim 1, wherein said inputting the operational data into a first fault identification model and a second fault identification model comprises:
respectively determining fault characteristics corresponding to the first fault type and fault characteristics corresponding to the second fault type;
inputting first operation data matched with the fault characteristics corresponding to the first fault type in the operation data into the first fault identification model;
and inputting second operation data matched with the fault characteristics corresponding to the second fault type in the operation data into the first fault identification model.
3. The method of identifying a fault in a spindle of a machine tool according to claim 1, wherein the model parameters include weight values;
the first fault recognition model and the pre-training model have the same weight value.
4. A training method of a fault recognition model for recognizing a fault of a spindle of a machine tool, the training method comprising:
training a first neural network by using first fault sample data corresponding to a first fault type of the machine tool spindle to obtain a first fault identification model for identifying the first fault type;
constructing a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model;
and training the pre-training model by using second fault sample data corresponding to a second fault type of the machine tool spindle to obtain a second fault identification model for identifying the second fault type.
5. The method of identifying a failure of a spindle of a machine tool according to claim 4, further comprising:
determining a fault signature corresponding to the first fault type;
performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain first fault sample data of fault characteristics corresponding to the first fault type;
and/or, determining a fault signature corresponding to the second fault type;
and performing order analysis and wavelet packet decomposition on the operation data of the machine tool spindle to obtain second fault sample data corresponding to the fault characteristics of the second fault type.
6. The method of identifying a fault in a spindle of a machine tool according to claim 4, wherein the model parameters include weight values;
the constructing of the pre-training model of the second fault recognition model according to the model parameters of the first fault recognition model comprises:
and configuring the weight value of the pre-training model according to the weight value of the first fault recognition model.
7. The method of identifying a failure of a spindle of a machine tool according to claim 4, further comprising:
training a second neural network by using the second fault sample data to obtain a verification model of the second fault identification model, wherein the second neural network and the pre-training model have the same network structure;
performing five-fold cross validation on the second fault identification model according to the validation model;
and determining the second fault identification model passing the verification as a final second fault identification model.
8. A failure recognition device for a spindle of a machine tool, comprising:
the acquisition module is used for acquiring the operating data of the machine tool spindle;
the input module is used for inputting the operation data into a first fault identification model and a second fault identification model; the first fault recognition model is obtained by training a neural network according to first fault sample data corresponding to a first fault type of the machine tool spindle, the second fault recognition model is obtained by training a pre-training model according to second fault sample data corresponding to a second type of fault of the machine tool spindle, and the pre-training model is constructed according to model parameters of the first fault recognition model;
and the identification module is used for carrying out fault identification on the machine tool spindle by using the first fault identification model and the second fault identification model.
9. A training apparatus for a fault recognition model for recognizing a fault of a spindle of a machine tool, comprising:
the first training module is used for training a first neural network by using first fault sample data corresponding to a first fault type of the machine tool spindle to obtain a first fault identification model for identifying the first fault type;
the building module is used for building a pre-training model of a second fault recognition model according to the model parameters of the first fault recognition model;
and the second training module is used for training the pre-training model by using second fault sample data corresponding to a second fault type of the machine tool spindle to obtain a second fault identification model for identifying the second fault type.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method of any of claims 1 to 7 when executing the computer program.
11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110895587.7A CN113762344A (en) | 2021-08-05 | 2021-08-05 | Fault identification method of machine tool spindle and fault identification model training method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110895587.7A CN113762344A (en) | 2021-08-05 | 2021-08-05 | Fault identification method of machine tool spindle and fault identification model training method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113762344A true CN113762344A (en) | 2021-12-07 |
Family
ID=78788700
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110895587.7A Pending CN113762344A (en) | 2021-08-05 | 2021-08-05 | Fault identification method of machine tool spindle and fault identification model training method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113762344A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114971252A (en) * | 2022-05-19 | 2022-08-30 | 浙江理工大学 | Operation and maintenance and fault pre-diagnosis system for textile equipment |
| CN115994308A (en) * | 2023-03-22 | 2023-04-21 | 中科航迈数控软件(深圳)有限公司 | Numerical control machine tool fault identification method, system, equipment and medium based on meta learning |
| CN116703284A (en) * | 2023-08-03 | 2023-09-05 | 八爪鱼人工智能科技(常熟)有限公司 | Fault identification method applied to refrigeration house management system and artificial intelligent server |
| CN117762086A (en) * | 2024-02-22 | 2024-03-26 | 东莞市微振科技有限公司 | machine tool parameter processing method and device, electronic equipment and readable storage medium |
| CN117762086B (en) * | 2024-02-22 | 2024-05-28 | 东莞市微振科技有限公司 | Machine tool parameter processing method and device, electronic equipment and readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560997A (en) * | 2020-12-29 | 2021-03-26 | 珠海拓芯科技有限公司 | Fault recognition model training method, fault recognition method and related device |
| CN113128561A (en) * | 2021-03-22 | 2021-07-16 | 南京航空航天大学 | Machine tool bearing fault diagnosis method |
-
2021
- 2021-08-05 CN CN202110895587.7A patent/CN113762344A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560997A (en) * | 2020-12-29 | 2021-03-26 | 珠海拓芯科技有限公司 | Fault recognition model training method, fault recognition method and related device |
| CN113128561A (en) * | 2021-03-22 | 2021-07-16 | 南京航空航天大学 | Machine tool bearing fault diagnosis method |
Non-Patent Citations (1)
| Title |
|---|
| 曹宁: "小样本下基于迁移学习的轴承状态识别方法", 噪声与振动控制, 31 October 2020 (2020-10-31), pages 1 - 7 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114971252A (en) * | 2022-05-19 | 2022-08-30 | 浙江理工大学 | Operation and maintenance and fault pre-diagnosis system for textile equipment |
| CN115994308A (en) * | 2023-03-22 | 2023-04-21 | 中科航迈数控软件(深圳)有限公司 | Numerical control machine tool fault identification method, system, equipment and medium based on meta learning |
| CN116703284A (en) * | 2023-08-03 | 2023-09-05 | 八爪鱼人工智能科技(常熟)有限公司 | Fault identification method applied to refrigeration house management system and artificial intelligent server |
| CN116703284B (en) * | 2023-08-03 | 2023-10-17 | 八爪鱼人工智能科技(常熟)有限公司 | Fault identification method applied to refrigeration house management system and artificial intelligent server |
| CN117762086A (en) * | 2024-02-22 | 2024-03-26 | 东莞市微振科技有限公司 | machine tool parameter processing method and device, electronic equipment and readable storage medium |
| CN117762086B (en) * | 2024-02-22 | 2024-05-28 | 东莞市微振科技有限公司 | Machine tool parameter processing method and device, electronic equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Theissler et al. | Predictive maintenance enabled by machine learning: Use cases and challenges in the automotive industry | |
| JP7105932B2 (en) | Anomaly detection using deep learning on time series data related to application information | |
| CN113762344A (en) | Fault identification method of machine tool spindle and fault identification model training method and device | |
| US10831577B2 (en) | Abnormality detection system, abnormality detection method, abnormality detection program, and method for generating learned model | |
| Wang et al. | Temporal convolutional network with soft thresholding and attention mechanism for machinery prognostics | |
| EP3139313B1 (en) | Anomaly detection system and method | |
| CN102520697B (en) | Onsite information preprocessing method of remote cooperative diagnosis | |
| CN113128561A (en) | Machine tool bearing fault diagnosis method | |
| Yu | A nonlinear probabilistic method and contribution analysis for machine condition monitoring | |
| CN112633339A (en) | Bearing fault intelligent diagnosis method, bearing fault intelligent diagnosis system, computer equipment and medium | |
| CN110426634B (en) | Method and equipment for predicting abnormity of driving system | |
| CN112393931A (en) | Detection method, detection device, electronic equipment and computer readable medium | |
| Cipollini et al. | Unsupervised deep learning for induction motor bearings monitoring | |
| Mathew et al. | Regression kernel for prognostics with support vector machines | |
| Zschech et al. | Prognostic model development with missing labels: a condition-based maintenance approach using machine learning | |
| Gao et al. | A process fault diagnosis method using multi‐time scale dynamic feature extraction based on convolutional neural network | |
| Nasser et al. | A hybrid of convolutional neural network and long short-term memory network approach to predictive maintenance | |
| KR20220062547A (en) | Sensor Agnostic Mechanical Mechanical Fault Identification | |
| CN116380445A (en) | Equipment state diagnosis method and related device based on vibration waveform | |
| CN115859077A (en) | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions | |
| Deac et al. | Vibration anomaly detection using deep neural network and convolutional neural network | |
| KR102306244B1 (en) | Method and apparatus for generating fault detecting model for device | |
| Melendez et al. | Self-supervised Multi-stage Estimation of Remaining Useful Life for Electric Drive Units | |
| CN115659271A (en) | Sensor abnormality detection method, model training method, system, device, and medium | |
| Techane et al. | Rotating machinery prognostics and application of machine learning algorithms: Use of deep learning with similarity index measure for health status prediction |
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 |