CN114310488B - Method for generating cutter fracture detection model, detection method, equipment and medium - Google Patents
Method for generating cutter fracture detection model, detection method, equipment and medium Download PDFInfo
- Publication number
- CN114310488B CN114310488B CN202111616505.7A CN202111616505A CN114310488B CN 114310488 B CN114310488 B CN 114310488B CN 202111616505 A CN202111616505 A CN 202111616505A CN 114310488 B CN114310488 B CN 114310488B
- Authority
- CN
- China
- Prior art keywords
- data
- processing data
- processing
- cutter
- acquiring
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 99
- 238000001514 detection method Methods 0.000 title claims abstract description 69
- 238000012545 processing Methods 0.000 claims abstract description 204
- 238000009826 distribution Methods 0.000 claims abstract description 70
- 208000010392 Bone Fractures Diseases 0.000 claims abstract description 52
- 206010017076 Fracture Diseases 0.000 claims abstract description 52
- 230000008569 process Effects 0.000 claims abstract description 46
- 238000012549 training Methods 0.000 claims abstract description 42
- 230000005856 abnormality Effects 0.000 claims abstract description 33
- 238000003754 machining Methods 0.000 claims description 65
- 230000005540 biological transmission Effects 0.000 claims description 29
- 238000005520 cutting process Methods 0.000 claims description 27
- 238000007781 pre-processing Methods 0.000 claims description 13
- 238000012216 screening Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 abstract description 4
- 238000013136 deep learning model Methods 0.000 abstract description 2
- 239000013598 vector Substances 0.000 description 11
- 230000006870 function Effects 0.000 description 10
- 238000003860 storage Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000001914 filtration Methods 0.000 description 5
- 238000013507 mapping Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 230000002411 adverse Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000003801 milling Methods 0.000 description 3
- 230000003139 buffering effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000007480 spreading Effects 0.000 description 2
- 238000003892 spreading Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013506 data mapping Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The application discloses a method, a device and a medium for generating a cutter fracture detection model, wherein the method comprises the following steps: acquiring a plurality of first processing data; constructing a training set according to a plurality of first processing data; training a preset distribution system model through a training set to obtain a cutter fracture detection model, wherein the cutter fracture detection model is used for obtaining the degree of abnormality of a cutter to be detected, and judging whether the cutter fracture risk exists or not according to the degree of abnormality and an abnormality threshold value. According to the application, only a plurality of times of normal processing data are needed to construct a training set to train a preset distribution system model, so that various complex process procedures can be dealt with; according to the application, a deep learning model is established without complex environment construction, so that the data processing efficiency is improved; the application is based on high mathematical abstraction, and can obtain a stable and effective model capable of accurately detecting the fracture of the cutter. In addition, the application can detect whether the cutter has fracture risk in real time in the processing process, and has higher timeliness.
Description
Technical Field
The present application relates to the field of tool detection, and in particular, to a method, a device, and a medium for generating a tool fracture detection model.
Background
The current-stage cutter fracture detection is divided into a direct method and an indirect method. The direct method measures the external dimension, the geometric shape and the like of the cutter through physical contact or laser detection, so that the judgment of cutter fracture is realized, and the contact type or laser cutter setting gauge and other products are representatively realized at the present stage. The direct method tool breakage detection is simple and accurate, but requires an additional breakage detection step for the tool after a section of processing is completed, which certainly increases the time cost of processing.
The indirect method is a method for assisting in indirectly judging the current machining state by using data generated during machining accompanied by machining, for example, an experienced operator can usually perceive whether the current equipment is operating normally through sound and vibration of the equipment. The indirect method has the advantages that the current processing state of the equipment can be reflected on processing data in real time, and the processing technology is deeply bound with a data source, so that the abnormality of the equipment can be distinguished in the processing process or when the processing is finished, and the additional steps and the processing time are not required to be added. However, because the data is related to the depth of machining, random noise, interference and other factors in the machining process of equipment can be truly reflected in the machining data, so that instability in pattern recognition is increased, and the application of indirect measurement-based tool monitoring in actual machining scenes is further limited.
The currently commonly used indirect method is to directly acquire data representing processing from a central control system of a machine, and the mode has low sampling rate, weak correlation degree between the data and the processing and insufficient resolution for fine processing, so that accurate cutter breakage judgment on the fine processing is difficult.
In addition, when constructing the break detection algorithm, a lot of equipment hardware costs, calculation time costs, data costs required for modeling, and the like are often required.
Disclosure of Invention
The application aims to overcome the defects of high cost and inaccurate judgment result of the broken cutter judgment in the machining process, particularly in the fine machining process in the prior art, and provides a method, a device and a medium for generating a cutter fracture detection model, wherein the method, the device and the medium can accurately perform the broken cutter judgment in the machining process, particularly in the fine machining process, while reducing the cost.
The application solves the technical problems by the following technical scheme:
the application provides a generation method of a cutter fracture detection model, which comprises the following steps:
acquiring a plurality of first machining data, wherein the first machining data represent machining data of a cutter for cutting a target workpiece under a preset cutting process;
constructing a training set according to the plurality of first processing data;
and training a preset distribution system model through the training set to obtain a cutter fracture detection model and an abnormality threshold, wherein the cutter fracture detection model is used for obtaining the abnormality of a cutter to be detected and judging whether the cutter to be detected has fracture risk according to the abnormality and the abnormality threshold.
Preferably, the step of constructing a training set according to the plurality of first processing data includes:
for each first processing data, preprocessing the first processing data to obtain second processing data, wherein the second processing data is formed by removing interference data in the first processing data;
and constructing a training set according to a plurality of the second processing data.
Preferably, the first processing data includes machine data;
when the first processing data is the machine data, the step of preprocessing the first processing data includes:
acquiring third machining data and corresponding spindle rotating speed data under a preset cutting process;
and acquiring a fluctuation range corresponding to the spindle rotation speed data, and taking the third processing data as second processing data when the fluctuation range is smaller than or equal to a preset fluctuation range.
Preferably, the first process data comprises sensor data;
when the first process data is the sensor data, the step of preprocessing the first process data to obtain second process data for each first process data includes:
for each first process data, the first process data is low pass filtered to obtain second process data.
Preferably, when the first processing data is the sensor data, the step of preprocessing the first processing data to obtain second processing data for each first processing data includes:
acquiring fourth machining data, wherein the fourth machining data are machining data of the cutter which does not cut a workpiece under the preset cutting process;
and for each first processing data, screening out target processing data according to the difference between the first processing data and the fourth processing data, and taking the target processing data as second processing data.
Preferably, the step of obtaining the plurality of first processing data further includes:
for each first processing data, acquiring a corresponding first energy distribution time sequence according to the first processing data;
the step of obtaining the fourth processing data further comprises the following steps:
acquiring a corresponding second energy distribution time sequence according to the fourth processing data;
the step of screening out target processing data according to the difference between the first processing data and the fourth processing data includes: and screening out a target frequency band according to the difference between the first energy distribution time sequence and the second energy distribution time sequence, and acquiring target processing data of the target frequency band.
Preferably, the step of acquiring the plurality of first processing data includes:
when the first processing data are machine data, acquiring a plurality of first processing data in a segmented transmission mode;
when the first processing data are sensor data, acquiring a plurality of first processing data in a real-time transmission mode or a windowed transmission mode; and/or the number of the groups of groups,
the preset distribution system model comprises a von mises fischer distribution system.
The application also provides a detection method of the cutter fracture, which comprises the following steps:
acquiring to-be-detected processing data of a to-be-detected tool, wherein the to-be-detected processing data represent processing data of a target workpiece cut by the to-be-detected tool under a preset cutting process;
inputting the processing data to be detected into a tool fracture detection model to obtain tool abnormality, wherein the tool fracture detection model is obtained according to the generation method of the tool fracture detection model;
and when the cutter abnormality degree is greater than or equal to an abnormality degree threshold, generating prompt information, wherein the prompt information is used for indicating that the cutter to be detected has fracture risk, and the abnormality degree threshold is a threshold generated according to the cutter fracture detection model.
The application also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the generation method of the tool breakage detection model or the detection method of the tool breakage when executing the computer program.
The present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of generating a tool breakage detection model as described above or the method of detecting a tool breakage as described above.
The application has the positive progress effects that: according to the application, a training set can be constructed by acquiring a plurality of first processing data, and a preset distribution system model is trained by the training set to acquire a cutter fracture detection model and an abnormality threshold value so as to detect whether a cutter to be detected has fracture risk. According to the application, the training set training preset distribution system model can be constructed only by the data of a plurality of times of normal processing, various complex process procedures can be dealt with, and for different processes, the corresponding cutter fracture detection model can be constructed by the data of a plurality of times of normal processing, so that large-scale spreading application is possible. According to the application, a deep learning model is established without complex environment construction, so that the data processing efficiency is improved, and the data processing cost is reduced. And the application is based on high mathematical abstraction, and can obtain a model which is stable, effective and capable of accurately detecting the fracture of the cutter.
The application can detect whether the cutter has fracture risk in real time in the cutter processing process, can monitor whether the cutter is broken in the processing process, does not need to wait until the processing is finished and then judges, and has higher timeliness.
Drawings
Fig. 1 is a flowchart of a method for generating a tool breakage detection model in embodiment 1 of the present application.
Fig. 2 is a flowchart of a first implementation of step 102 in embodiment 1 of the present application.
Fig. 3 is a flowchart of a second implementation of step 102 in embodiment 1 of the present application.
Fig. 4 is a flowchart of a third implementation of step 102 in embodiment 1 of the present application.
Fig. 5 is a flowchart of a method for detecting tool breakage in embodiment 2 of the present application.
Fig. 6 is a schematic block diagram of an electronic device in embodiment 3 of the present application.
Detailed Description
For ease of understanding, the terms commonly found in the examples are explained below:
the terms "having," "may have," "including," or "may include," as used herein, indicate the presence of a corresponding function, operation, element, etc. of the disclosure, and are not limited by the presence of other one or more functions, operations, elements, etc. Furthermore, it should be understood that the terms "comprises" or "comprising," as used herein, are intended to specify the presence of stated features, integers, steps, operations, elements, components, or groups thereof, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or groups thereof.
The term "a or B", "at least one of a and/or B" or "one or more of a and/or B" as used herein includes any and all combinations of words listed therewith. For example, "a or B", "at least one of a and B" or "at least one of a or B" means (1) including at least one a, (2) including at least one B, or (3) including both at least one a and at least one B.
The first, second, etc. descriptions appearing in the embodiments of the present application are merely for purposes of illustration and distinguishing between the objects of description, and are not intended to be limiting in any way, nor are the number of devices specifically defined in the embodiments of the present application. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of the present disclosure.
The expression "configure (set up)" as used herein may be replaced with "adapted to", "having the ability", "designed to", "adapted to", "formed to" or "capable" as the case may be. The term "configured to (set up)" does not necessarily mean "specifically designed to" at the hardware level. Conversely, the expression "configure …" may mean that the apparatus is "capable" in some cases along with other devices or components. For example, the "processor configured to (set) execute A, B and C" may be a dedicated processor such as an embedded processor for executing corresponding operations, or a general-purpose processor such as a Central Processing Unit (CPU) or an Application Processor (AP) capable of executing corresponding operations by running one or more software programs stored in a storage device.
The application is further illustrated by means of the following examples, which are not intended to limit the scope of the application.
Example 1
The embodiment provides a method for generating a tool fracture detection model, as shown in fig. 1, which comprises the following steps:
step 101, acquiring a plurality of first processing data.
The first machining data represents machining data of a tool cutting a target workpiece under a preset cutting process.
For example, each processing is pretreated to obtain a data sequence s after reserving the processing section i S (N, i|p, T), where P, T are the machining process number and the machining tool number, respectively, N isRandom noise, i, represents the ith process. Set F agg For an aggregate function (aggregate function represents a function mapping an object of complex elements, such as a vector, a matrix, etc., to a single real number value) (functional represents a function used to generate the aggregate function, the generation control is performed by a hyper-parameter in the course of generating the aggregate function), F is F agg Or any custom data mapping method that can achieve the effect of function aggregation. Construction of mapping operator F for extracting features from data features map ＝(f 1 ,f 2 ,...,f n ) Each processing sequence is mapped to become a vector with length n, namely x i ＝F map (s i )，x i ∈R n . The problem of comparing the original processing sequences at this time translates into a pattern creation and comparison problem in n-dimensional space. In this embodiment, n is selected as small as possible on the premise of representing the processing characteristics, so as to improve the efficiency of data processing.
The first processing data can be specifically divided into two types of data, wherein one type is that data representing processing, such as power, load and the like, are directly obtained from a central control system of the machine station; the other is to acquire data associated with machining, such as vibration, acoustic emission sensors, etc., by externally adding sensors. The two methods are characterized in that: the former is usually low in sampling rate, weak in association degree of data and processing, and insufficient in resolution for fine processing; the latter is purer in data source, less in invalid interference, but larger in data scale and relatively lower in signal-to-noise ratio.
Therefore, in this embodiment, the preset cutting process is specifically a rough machining process, or a process with low refinement requirement may use machine data to improve the convenience of data acquisition, and the process with high refinement requirement is preferably sensor data to improve the accuracy of data acquisition.
In this embodiment, the first processing data may be obtained by means of real-time transmission, pane transmission, segment transmission, and the like. The real-time transmission indicates that the data acquisition device (such as a machine station and a sensor) can send a piece of data to the data processing module in real time for subsequent data processing when the data acquisition device acquires the piece of data newly; the windowed transmission indicates that each new piece of data is acquired, the data is temporarily stored in a preset pane, the earliest piece of data in the pane is deleted, and when the preset new data number (step length) is acquired, data transmission is carried out once by taking the pane as a unit; the segmented transmission means that the acquired new data continuously enter a buffer memory, and when the end of one segment of processing is monitored, unified transmission is carried out. The buffer storage requirement of the three transmission modes for the acquisition module is gradually increased from front to back.
The data collection process can be configured in advance, for example, a general configuration (for example, the number of data transmitted in unit time is set, generally, the larger the number of data transmitted in unit time is, the larger the difference between the actual collection efficiency and the theoretical collection efficiency is, so that the data transmitted in unit time needs to be simplified as much as possible in combination with the transmission efficiency), and for example, a special configuration is performed in a specific transmission mode, where the special configuration is set according to personalized parameters corresponding to different data transmission modes, for example, parameters such as pane length, step length and the like required to be preset during the pane transmission.
In this embodiment, for different types of first processing data, different data transmission modes may be set, and specifically, step 101 may include the following steps:
when the first processing data are machine data, acquiring a plurality of first processing data in a segmented transmission mode;
and when the first processing data are sensor data, acquiring a plurality of first processing data in a real-time transmission mode or a windowed transmission mode.
In this embodiment, when the first processing data is machine data, since the machine data is directly utilized, the acquisition frequency of the machine data is low, the requirement on buffering is not large, so that the creation of a subsequent mode is convenient, therefore, a segmented transmission mode can be adopted at the acquisition end under the condition of avoiding adverse effects on buffering, so that the subsequent means for processing the first processing data can be flexibly added, for example, only current data can be acquired during real-time transmission, data average in a period of time can be acquired during pane transmission, and more detailed information such as average and fluctuation range of the processing section data can be simultaneously calculated during segmented transmission. When the first processing data is sensor data, the data quantity generated in unit time is large due to high sensor data acquisition frequency, so that real-time transmission or pane transmission with small influence on the cache can be adopted, and adverse influence on the cache is reduced. By the method, adverse effects on caching can be reduced, and data processing efficiency is improved.
Step 102, constructing a training set according to a plurality of first processing data.
If a positive integer m not less than n is set, standard processing is performed m times to generate m data sequences S model ＝{s 1 ,s 2 ,…,s m }. Considering the cost factor, m can be as small as possible under the condition that n is not less. Using the mapping operator F map For S model The elements in the model are mapped one by one to obtain a feature set X of the model model ＝{x 1 ,x 2 ,…,x m }. Definition m= (x 1 ,x 2 ,…,x m ) T For processing the data model, M is known as R m×n I.e. the training set.
Since there may be some interference data, such as random noise, in the directly obtained first machining data, which is not related to the judgment of the cutting edge, as shown in fig. 2, step 102 may further include the following steps to remove the interference data:
step 1021, for each first processing data, preprocessing the first processing data to obtain second processing data.
The second processing data is formed by removing interference data in the first processing data.
Step 1022, constructing a training set according to the plurality of second processing data.
In this embodiment, the first processing data may be preprocessed to generate the second processing data, so that the portion of the first processing data, which belongs to random noise, may be eliminated as much as possible, so as to improve the signal-to-noise ratio of the data, and facilitate subsequent creation of a more effective model.
The data acquired in step 101 may be derived from the data of the machine itself or from the data of the external sensor. The data sources are different, and the corresponding preprocessing flows are also different. Specifically, when the first processing data is machine data, as shown in fig. 3, in step 102, the step of preprocessing the first processing data includes:
step 1121, obtaining third machining data under a preset cutting process and corresponding spindle rotation speed data;
and 1122, acquiring a fluctuation range corresponding to the spindle rotation speed data, and taking the third processing data as the second processing data when the fluctuation range is smaller than or equal to a preset fluctuation range.
In this embodiment, for the machine self data, since the machine self data and the actual working condition are indirectly represented by some physical quantities, these physical quantities are not generally in one-to-one correspondence with the actual working condition. For example, it is a common way to characterize the magnitude of the machining cutting force by the machine load, and the machine load increases as the machining cutting force increases, but conversely, the machine can also generate a significant load in the non-machining state, such as when performing a tool change or the like. The model creation process based on the machine data is performed based on the machine data, and therefore, in the preprocessing step, the data generated in the similar non-machining state needs to be screened and excluded. Because the spindle always maintains a stable rotation speed in the machining state, the present embodiment adopts a method of synchronously checking the rotation speed of the spindle to perform preprocessing of removing the non-machining load. The specific method comprises the steps of designing a certain time delay, synchronously reading the load and spindle rotation speed data of the machine, calculating the fluctuation range of the spindle rotation speed, and reading the load data of the machine at the moment when the fluctuation range is smaller than a preset value, or else, discarding the load data. By the method, interference data can be effectively eliminated, and more accurate second processing data can be obtained. After the machine data is preprocessed, although machining background noise still exists, the influence on the subsequent modeling step is small, so that a model capable of accurately detecting whether a cutter has fracture risk can be built based on the preprocessed machine data.
When the first processing data is sensor data, step 102 specifically includes:
for each first process data, the first process data is low pass filtered to obtain second process data.
Specifically, the first processing data can be framed in a preset pane, and the vibration data is subjected to moderate low-pass filtering to obtain second processing data so as to eliminate interference data by taking the pane as a unit, and the sensor data is only sensitive to the measured physical quantity, so that when the machine performs certain non-processing operations, some larger data interference is not introduced as the machine data, and therefore, the operation can be simplified in a low-pass filtering mode so as to improve the effectiveness of training data acquisition.
In a specific implementation, as shown in fig. 4, step 102 may specifically include the following steps:
step 1221, obtaining fourth processing data.
The fourth processing data is the processing data of the cutter without cutting the workpiece under the preset cutting process.
Step 1222, for each first processing data, screening out target processing data according to the difference between the first processing data and the fourth processing data, and using the target processing data as the second processing data.
In this embodiment, a set of typical cutting process combinations (i.e., preset cutting processes) is designed as standard machining to obtain the first machining data, wherein the typical cutting process should be able to include most of the machining scenes to be monitored, such as side milling, end milling, etc. for the forming and milling stage. After the standard processing is executed for one time, a section of idle running under the same rotating speed is executed to obtain fourth processing data, the fourth processing data is used as standard checking process data, scoring is made according to the difference between the first processing data and the fourth processing data, and data in normal processing can be automatically screened out according to the scoring, so that the effectiveness of data acquisition is further improved.
In a preferred implementation, step 102 may further include the steps of: and for each first processing data, acquiring a corresponding first energy distribution time sequence according to the first processing data.
Preferably, the first processing data may be first subjected to low-pass filtering to exclude interference data as much as possible, and the corresponding first energy distribution time sequence is obtained according to the low-pass filtered data.
Step 1121 may be followed by the further step of: and acquiring a corresponding second energy distribution time sequence according to the fourth processing data.
Preferably, the fourth processing data may be first subjected to low-pass filtering to exclude interference data as much as possible, and the corresponding second energy distribution time sequence is obtained according to the data after the low-pass filtering.
Step 1122 may specifically include the steps of: and screening out a target frequency band according to the difference between the first energy distribution time sequence and the second energy distribution time sequence, and acquiring target processing data of the target frequency band.
The sensor data is not very intuitive to correspond to the conditions in real-world processing. Taking a vibration sensor as an example, when the machining cutting force increases, it can be observed that the amplitude of the vibration increases. However, it is difficult to build an effective data model only by vibration amplitude (amplitude fluctuations are affected by the process but they themselves have some randomness.
In this embodiment, the sensor data is preprocessed by converting to energy distribution. Taking vibration as an example, the amplitude of vibration data and the fluctuation thereof always have certain randomness, however, the energy distribution is relatively stable, the energy distribution of the vibration data is converted, and the converted data is the energy distribution intensity values of different frequency sections. Through the mode of converting the data into the energy distribution time sequence, the data can be more visual, and the data can be conveniently analyzed. Specifically, according to the energy distribution time sequence, the difference between each energy frequency band under normal processing and when the energy frequency band runs in the air can be scored, and the frequency band belonging to the processing energy is automatically screened out according to the score to serve as target processing data.
And 103, training a preset distribution system model through a training set to obtain a cutter fracture detection model and an abnormality threshold.
The tool fracture detection model is used for acquiring the degree of abnormality of the tool to be detected and judging whether the tool to be detected has fracture risk according to the degree of abnormality and the threshold value of the degree of abnormality.
In this example, the pre-set distribution system model specifically employs the von mises fischer distribution system, it being understood that other models for performing distribution statistics may be employed in other embodiments.
The principle of the training process of the preset distribution system model is explained below by taking von mises fischer distribution system as an example:
conventional methods generally require both positive and negative sample data to be obtained to train the model. In the case of the broken knife detection, if the broken knife does not occur, the obtained data are similar (because the obtained data are data similar to those of standard processing), but if the broken knife occurs, the obtained data are different (because the cause of the broken knife may be various, the time and the scene of the broken knife occur are various), and in the actual operation process, it is difficult to collect the comprehensive negative sample. Therefore, if the breaking detection model is trained based on the conventional manner, on one hand, the cost of collecting the training set is too high, and on the other hand, since the collected negative sample cannot cover the occurrence of each breaking, the trained breaking detection model cannot detect the occurrence of the breaking under the condition that another breaking occurs, so that the trained model is difficult to accurately detect the occurrence of the breaking.
In contrast to the conventional method of constructing a machine learning model by using both positive and negative samples during tool detection, the present embodiment only needs to collect positive samples, and by inputting the training set, i.e., the data during standard processing (i.e., the positive samples), into the von mises fisher distribution system, the standard processing data can be converted into statistical data distribution, i.e., the tool breakage detection model, and the maximum error rate allowed under the data distribution, i.e., the anomaly threshold.
When the detection is actually carried out, only the processing data to be detected is required to be input into the cutter fracture detection model, and the data distribution can be obtained. And judging whether the data distribution is in the data distribution range of the standard machining data, if the data distribution is not in the data distribution range, the machining data to be detected and the data distribution of the standard machining data are indicated to have larger difference, so that the risk of cutter breakage possibly exists, otherwise, if the data distribution is in the standard machining data distribution range, the distribution of the machining data to be detected and the standard machining data is indicated to be approximate, and the risk of cutter breakage does not exist.
By means of the model training mode in the embodiment, on one hand, the collection cost of the training set can be reduced (only positive samples are required to be collected, and because the positive samples are data in standard processing, the data are similar, and therefore a plurality of positive samples are not required to be collected), on the other hand, the principle of model implementation is that data distribution of data to be detected is compared with data distribution in standard processing, the situation that broken cutters are not detected due to the fact that broken cutters occur and different occurring scenes can be avoided, and the accuracy of broken cutter detection can be improved.
Since probability statistics is allowed to have certain errors, the abnormality degree threshold a th It is used to represent such errors, which in particular represent an upper limit allowing the data distribution of the data to be detected to be out of the range of the distribution of the standard data, if the corresponding degree of abnormality obtained for the data to be detected exceeds this upper line, it means that its corresponding data distribution is largely out of the range of the distribution of the standard data, and therefore the tool has a risk of breakage.
Wherein the abnormality degree threshold value a th Has a functional relationship with the test level a, in particular as follows:
the test level a can be set by oneself, before training, the value of the test level a can be set according to the actual requirement, when the value of the test level a is set to be higher, the trained abnormality threshold value a is correspondingly used th The model trained to detect the broken blade is lower, i.e., more stringent. Conversely, when the value of the inspection level a is set to be lower, the trained abnormality threshold a is correspondingly trained th The higher the model trained would be, the more relaxed the detection of a broken knife.
Specifically, in this embodiment, the training set, i.e., the matrix M, is input to the von mises fischer distribution system, and the following procedure is performed by the von mises fischer distribution system: for all M of M i,: (i.e {1,2, …, m }) to obtain m i,: The average unit vector μ of M is obtained, and a is used i,: ＝1-μ T m i,: Calculating the abnormality degree of each sample in M to obtain a set { a } obeying the chi-square distribution 1,: ,a 2,: ,…,a m,: }～χ 2 (m mo ,s mo ) Wherein m is mo Sum s mo Is a moment estimation of two parameters of the chi-square distribution, wherein one parameter is used for estimating the direction of data distribution, the other parameter is used for estimating the concentration of the data distribution, and a threshold value a of the degree of anomaly is calculated according to the designed inspection level a th 。
In this embodiment, a training set may be constructed according to a plurality of first processing data, specifically, feature extraction parameters may be configured, after feature extraction is performed on the processing data, the processing data are arranged into vectors with equal length, the training set is constructed according to vectors processed normally for a plurality of times, in a von mises fischer distribution system, the outlier of each vector is calculated, and based on the chi-square distribution of the outlier of all vectors, an outlier threshold may be further calculated based on the chi-square distribution, so that for each subsequent processing, a vector with equal length may be obtained through the same processing procedure as the data and based on the tool breakage detection model, and whether the tool to be detected has a risk of breakage or not may be effectively determined according to the calculated outlier of the vector and the outlier threshold.
According to the embodiment, the training set can be constructed by acquiring a plurality of first processing data, and the preset distribution system model is trained through the training set to acquire the cutter fracture detection model and the abnormality threshold so as to detect whether the cutter to be detected has fracture risk. The machining recognition accuracy of this embodiment is high: because in general, the detection of the fracture of the cutter by the indirect method is realized based on the machine data, the method has good performance only in the process of relatively large cutting amount, and the data generated along with the small cutting amount cannot be distinguished even from blank running, so that the identification is difficult; on-machine detection based on sensor data generally uses Root Mean Square (RMS) direct evaluation processing of vibration data, ignores information of a frequency domain part, and has an undesirable effect when applied to more complex processes. In the embodiment, for the process with larger cutting amount, the machine data is preferably used for identification; for machining with smaller cutting quantity, the data with higher signal-to-noise ratio can be obtained in a narrower frequency domain segment by the sensor and matched data transformation, so that precise identification of fine machining is realized. Thereby improving the accuracy of model training.
The data acquisition cost of the embodiment is low, the pattern recognition can be classified as a classification problem in the machine learning, and the conventional classification training mode training stable model needs a large number of training samples as classification samples, however, in actual machining, the situation of cutter breakage is extremely rare compared with normal machining. In other words, to obtain a stable and reliable model, a lot of time, manpower and experimental costs are required to simulate the breaking in each case, and this solution requires extremely high costs. The training set can be constructed by training the preset distribution system model only by using a plurality of times of normal processing data, various complex process procedures can be dealt with, and for different processes, a corresponding cutter fracture detection model can be constructed by using a plurality of times of normal processing data, so that large-scale spreading application is possible.
The method is based on high mathematical abstraction, high in stability, complex in environment construction due to traditional machine learning or deep learning, high in requirement on equipment hardware performance, and extra hardware cost is increased because operations are processed by a special server even if the model capacity is large and the calculation is complex. The model building of the embodiment fully combines the data characteristics of the cutter processing scene, creates corresponding distribution and inspection by taking high-probability statistics as a technical means, has low requirement on hardware computing performance (can realize basic vector operation), has high operation speed and low cost, and is easy to deploy.
Example 2
The embodiment provides a method for detecting cutter fracture, as shown in fig. 5, the method comprises the following steps:
step 201, obtaining processing data to be detected of a tool to be detected.
The machining data to be detected represents machining data of a target workpiece to be cut by the tool to be detected under the preset cutting process described in embodiment 1.
Step 202, inputting the processing data to be detected into a tool fracture detection model to obtain the tool abnormality.
The tool breakage detection model is a model obtained according to the method of generating the tool breakage detection model in embodiment 1;
and 203, generating prompt information when the degree of abnormality of the cutter is greater than or equal to the threshold value of the degree of abnormality.
The prompt information is used for indicating that the cutter to be detected has fracture risk, and the abnormality threshold is a threshold generated according to the cutter fracture detection model.
In this embodiment, the process of acquiring the processing data to be detected and the process of processing the processing data to be detected can refer to the process of acquiring and processing the first processing data in embodiment 1, and will not be described herein.
For example, for each new process, the process data to be detected, i.e. a new data sequence to be s, can be obtained new (ii) via the same mapping operation as in step 101 (refer to the mapping operator F constructed in step 101) map ＝(f 1 ,f 2 ,...,f n ) Means of (c) before being fed to the tool fractureThe vector x of the corresponding characteristic value can be obtained after the model is detected new Normalized toCorresponding to an abnormality of +.> When a is new ≤a th If the current machining mode is the same as the previous m standard machining modes, the current machining mode is considered to be greatly different from the standard machining mode. I.e. to indicate a risk of breakage.
In this embodiment, the detection principle is to compare the data distribution of the data to be detected with the data distribution of the standard processing trained by the tool fracture detection model, so that even if there are different reasons or different conditions of the tool fracture, whether the tool to be detected is fractured or not can be accurately detected based on the difference of the data distribution, and the accuracy of the tool fracture detection can be improved.
According to the embodiment, whether the cutter has the risk of fracture or not can be detected in real time in the cutter machining process, the cutter does not need to wait until the machining is finished, for example, the steps 201-203 are synchronously executed in the window data collecting process, whether the cutter is broken or not can be detected in the machining process, the cutter does not need to wait until the machining is finished, and the timeliness is higher.
According to the embodiment, based on the cutter fracture detection model, whether the cutter has fracture risk can be accurately detected, for different processing technologies, the corresponding cutter fracture detection model can be obtained according to the process of training a small amount of data, the cutter fracture detection cost is reduced, and the cutter fracture detection efficiency is also improved.
Example 3
The present embodiment provides an electronic device, which may be expressed in the form of a computing device (for example, may be a server device), including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor may implement the method for generating the tool breakage detection model in embodiment 1 or the method for detecting tool breakage in embodiment 2 when executing the computer program.
Fig. 6 shows a schematic diagram of the hardware structure of the present embodiment, and as shown in fig. 6, the electronic device 9 specifically includes:
at least one processor 91, at least one memory 92, and a bus 93 for connecting the different system components (including the processor 91 and the memory 92), wherein:
the bus 93 includes a data bus, an address bus, and a control bus.
The memory 92 includes volatile memory such as Random Access Memory (RAM) 921 and/or cache memory 922, and may further include Read Only Memory (ROM) 923.
Memory 92 also includes a program/utility 925 having a set (at least one) of program modules 924, such program modules 924 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.
The processor 91 executes various functional applications and data processing, such as the generation method of the tool breakage detection model in embodiment 1 of the present application or the detection method of the tool breakage in embodiment 2, by running the computer program stored in the memory 92.
The electronic device 9 may further communicate with one or more external devices 94 (e.g., keyboard, pointing device, etc.). Such communication may occur through an input/output (I/O) interface 95. Also, the electronic device 9 may communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet, through a network adapter 96. The network adapter 96 communicates with other modules of the electronic device 9 via the bus 93. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in connection with the electronic device 9, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, data backup storage systems, and the like.
It should be noted that although several units/modules or sub-units/modules of an electronic device are mentioned in the above detailed description, such a division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more units/modules described above may be embodied in one unit/module in accordance with embodiments of the present application. Conversely, the features and functions of one unit/module described above may be further divided into ones that are embodied by a plurality of units/modules.
Example 4
The present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of generating a tool breakage detection model in embodiment 1 or the method of detecting tool breakage in embodiment 2.
More specifically, among others, readable storage media may be employed including, but not limited to: portable disk, 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 embodiment, the application may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the method of generating a tool breakage detection model in example 1 or the method of detecting a tool breakage in example 2, when the program product is run on the terminal device.
Wherein the program code for carrying out the application may be written in any combination of one or more programming languages, which program code may execute entirely on the user device, partly on the user device, as a stand-alone software package, partly on the user device and partly on the remote device or entirely on the remote device.
While specific embodiments of the application have been described above, it will be appreciated by those skilled in the art that this is by way of example only, and the scope of the application 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 principles and spirit of the application, but such changes and modifications fall within the scope of the application.
Claims (5)
1. A method for generating a tool breakage detection model, the method comprising the steps of:
acquiring a plurality of first processing data, wherein the first processing data represent processing data of a cutter for cutting a target workpiece under a preset cutting process, and the first processing data comprise sensor data; the sensor data are preprocessed in a mode of converting the sensor data into energy distribution;
constructing a training set according to the plurality of first processing data;
training a preset distribution system model through the training set to obtain a cutter fracture detection model and an abnormality threshold, wherein the cutter fracture detection model is used for obtaining the abnormality of a cutter to be detected and judging whether the cutter to be detected has fracture risk according to the abnormality and the abnormality threshold;
the step of constructing a training set according to the plurality of first processing data comprises the following steps: for each first processing data, preprocessing the first processing data to obtain second processing data, wherein the second processing data is formed by removing interference data in the first processing data; constructing a training set according to a plurality of second processing data;
the step of obtaining a plurality of first processing data includes:
when the first processing data are machine data, acquiring a plurality of first processing data in a segmented transmission mode;
when the first processing data are sensor data, acquiring a plurality of first processing data in a real-time transmission mode or a windowed transmission mode; and/or the number of the groups of groups,
the preset distribution system model comprises a von mises fischer distribution system.
2. The method of generating a tool breakage detection model according to claim 1, wherein the first machining data includes machine data;
when the first processing data is the machine data, the step of preprocessing the first processing data includes:
acquiring third machining data and corresponding spindle rotating speed data under a preset cutting process;
and acquiring a fluctuation range corresponding to the spindle rotation speed data, and taking the third processing data as second processing data when the fluctuation range is smaller than or equal to a preset fluctuation range.
3. The method of generating a tool breakage detection model according to claim 1, wherein the first machining data includes sensor data;
when the first process data is the sensor data, the step of preprocessing the first process data to obtain second process data for each first process data includes:
for each first process data, the first process data is low pass filtered to obtain second process data.
4. The method of generating a tool breakage detection model according to claim 1, wherein when the first processing data is sensor data, the step of preprocessing the first processing data to obtain second processing data for each first processing data includes:
acquiring fourth machining data, wherein the fourth machining data are machining data of the cutter which does not cut a workpiece under the preset cutting process;
and for each first processing data, screening out target processing data according to the difference between the first processing data and the fourth processing data, and taking the target processing data as second processing data.
5. The method of generating a tool breakage detection model according to claim 4, wherein the step of acquiring a plurality of first machining data further comprises:
for each first processing data, acquiring a corresponding first energy distribution time sequence according to the first processing data;
the step of obtaining the fourth processing data further comprises the following steps:
acquiring a corresponding second energy distribution time sequence according to the fourth processing data;
the step of screening out target processing data according to the difference between the first processing data and the fourth processing data includes: and screening out a target frequency band according to the difference between the first energy distribution time sequence and the second energy distribution time sequence, and acquiring target processing data of the target frequency band.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111616505.7A CN114310488B (en) | 2021-12-27 | 2021-12-27 | Method for generating cutter fracture detection model, detection method, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111616505.7A CN114310488B (en) | 2021-12-27 | 2021-12-27 | Method for generating cutter fracture detection model, detection method, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114310488A CN114310488A (en) | 2022-04-12 |
| CN114310488B true CN114310488B (en) | 2023-10-27 |
Family
ID=81015887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111616505.7A Active CN114310488B (en) | 2021-12-27 | 2021-12-27 | Method for generating cutter fracture detection model, detection method, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114310488B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH04311206A (en) * | 1991-04-09 | 1992-11-04 | Agency Of Ind Science & Technol | Control method for avoiding abnormality in numerically controlled machine tool |
| CN101508035A (en) * | 2008-11-19 | 2009-08-19 | 深圳市大族激光科技股份有限公司 | Breaking monitoring method and system of machine-shaping device |
| CN102091972A (en) * | 2010-12-28 | 2011-06-15 | 华中科技大学 | Numerical control machine tool wear monitoring method |
| CN102179728A (en) * | 2011-03-14 | 2011-09-14 | 上海师范大学 | Device for intelligently detecting abrasion of numerical control cutting tool |
| CN106647629A (en) * | 2016-09-22 | 2017-05-10 | 华中科技大学 | Cutter breakage detection method based on internal data of numerical control system |
| CN110549162A (en) * | 2018-06-01 | 2019-12-10 | 日立汽车系统（中国）有限公司 | Cutting abnormality detection method and cutting abnormality detection device |
| CN112207631A (en) * | 2020-09-21 | 2021-01-12 | 深圳市玄羽科技有限公司 | Method for generating tool detection model, method, system, device and medium for detecting tool detection model |
| CN112257545A (en) * | 2020-10-19 | 2021-01-22 | 安徽领云物联科技有限公司 | Violation real-time monitoring and analyzing method and device and storage medium |
| CN113199304A (en) * | 2021-04-27 | 2021-08-03 | 中国科学院合肥物质科学研究院 | Tool changing monitoring method based on extended Kalman filtering and cutting force model |
| CN113752089A (en) * | 2021-10-19 | 2021-12-07 | 山东农业大学 | Cutter state monitoring method based on singularity Leersian index |
-
2021
- 2021-12-27 CN CN202111616505.7A patent/CN114310488B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH04311206A (en) * | 1991-04-09 | 1992-11-04 | Agency Of Ind Science & Technol | Control method for avoiding abnormality in numerically controlled machine tool |
| CN101508035A (en) * | 2008-11-19 | 2009-08-19 | 深圳市大族激光科技股份有限公司 | Breaking monitoring method and system of machine-shaping device |
| CN102091972A (en) * | 2010-12-28 | 2011-06-15 | 华中科技大学 | Numerical control machine tool wear monitoring method |
| CN102179728A (en) * | 2011-03-14 | 2011-09-14 | 上海师范大学 | Device for intelligently detecting abrasion of numerical control cutting tool |
| CN106647629A (en) * | 2016-09-22 | 2017-05-10 | 华中科技大学 | Cutter breakage detection method based on internal data of numerical control system |
| CN110549162A (en) * | 2018-06-01 | 2019-12-10 | 日立汽车系统（中国）有限公司 | Cutting abnormality detection method and cutting abnormality detection device |
| CN112207631A (en) * | 2020-09-21 | 2021-01-12 | 深圳市玄羽科技有限公司 | Method for generating tool detection model, method, system, device and medium for detecting tool detection model |
| CN112257545A (en) * | 2020-10-19 | 2021-01-22 | 安徽领云物联科技有限公司 | Violation real-time monitoring and analyzing method and device and storage medium |
| CN113199304A (en) * | 2021-04-27 | 2021-08-03 | 中国科学院合肥物质科学研究院 | Tool changing monitoring method based on extended Kalman filtering and cutting force model |
| CN113752089A (en) * | 2021-10-19 | 2021-12-07 | 山东农业大学 | Cutter state monitoring method based on singularity Leersian index |
Non-Patent Citations (1)
| Title |
|---|
| 吴陈编著.模式识别.《模式识别》.2020, * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114310488A (en) | 2022-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111459700B (en) | Equipment fault diagnosis method, diagnosis device, diagnosis equipment and storage medium | |
| JP6450858B2 (en) | Equipment management apparatus and method | |
| TWI422460B (en) | Tool nose detection method for cutting machine tool | |
| CN112207631B (en) | Method for generating tool detection model, method, system, device and medium for detecting tool detection model | |
| CN110653661A (en) | Cutter state monitoring and identifying method based on signal fusion and multi-fractal spectrum algorithm | |
| CN108520125A (en) | A kind of method and system of prediction cutting-tool wear state | |
| CN110346130A (en) | A kind of boring flutter detection method based on empirical mode decomposition and time-frequency multiple features | |
| US20220187164A1 (en) | Tool condition monitoring system | |
| CN112711850A (en) | Unit online monitoring method based on big data | |
| CN112067701A (en) | Fan blade remote auscultation method based on acoustic diagnosis | |
| CN114310488B (en) | Method for generating cutter fracture detection model, detection method, equipment and medium | |
| CN114227382A (en) | Cutter damage monitoring system and method based on novel capsule network | |
| CN116690689B (en) | Automatic cutting production line running state detection method | |
| CN114002332A (en) | Structural damage monitoring and early warning method and structural integrity digital twinning system | |
| CN113033639A (en) | Training method of abnormal data detection model, electronic device and storage medium | |
| US20230281186A1 (en) | Explainable anomaly detection for categorical sensor data | |
| CN111993158B (en) | Method for generating tool detection model, method, system, device and medium for detecting tool detection model | |
| CN113504302B (en) | Method and system for monitoring fan blade state, electronic equipment and storage medium | |
| CN109632081A (en) | Vibration of wind generating set feature extraction and otherness sentence method for distinguishing | |
| CN116150666B (en) | Energy storage system fault detection method and device and intelligent terminal | |
| US20230281478A1 (en) | Condition monitoring of an electric power converter | |
| CN114862283B (en) | Quality inspection method and quality inspection device for machine tool machining | |
| WO2024075567A1 (en) | Diagnostic system, information processing device, diagnostic method, and program | |
| CN113269222A (en) | Characteristic value selection method, detection method and system for detecting lifting equipment | |
| CN116861313B (en) | Kalman filtering working condition identification method and system based on vibration energy trend |
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 |