CN116628611A - Visual analysis method and system for association of abnormal modes of machine tool operation data - Google Patents
Visual analysis method and system for association of abnormal modes of machine tool operation data Download PDFInfo
- Publication number
- CN116628611A CN116628611A CN202310585965.0A CN202310585965A CN116628611A CN 116628611 A CN116628611 A CN 116628611A CN 202310585965 A CN202310585965 A CN 202310585965A CN 116628611 A CN116628611 A CN 116628611A
- Authority
- CN
- China
- Prior art keywords
- machine tool
- parameter
- data
- association
- tool operation
- 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
- 238000004458 analytical method Methods 0.000 title claims abstract description 70
- 230000000007 visual effect Effects 0.000 title claims abstract description 60
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 49
- 239000011159 matrix material Substances 0.000 claims abstract description 46
- 238000000034 method Methods 0.000 claims abstract description 34
- 238000005065 mining Methods 0.000 claims abstract description 29
- 230000002452 interceptive effect Effects 0.000 claims abstract description 15
- 238000007621 cluster analysis Methods 0.000 claims abstract description 13
- 238000013507 mapping Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 14
- 230000003993 interaction Effects 0.000 claims description 7
- 238000002759 z-score normalization Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000007405 data analysis Methods 0.000 claims description 4
- 230000005856 abnormality Effects 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 description 12
- 230000008859 change Effects 0.000 description 7
- 239000003086 colorant Substances 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000012098 association analyses Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/40—Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
-
- 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 invention discloses a machine tool operation data abnormal mode association visual analysis method and system, which are characterized in that symbol characteristics are respectively extracted from a sliding window and a time point of original machine tool operation data, multi-parameter class coding is carried out, a multi-parameter coding matrix sequence is obtained, multi-parameter class coding combinations of each column are extracted as data samples, cluster analysis and association rule mining are carried out, the machine tool operation mode is extracted according to the cluster analysis, and a parameter state association mode is extracted according to an association rule set obtained by the association rule mining; the interactive visual analysis system is constructed, visual analysis mapping of multi-parameter type codes, machine tool operation modes and parameter state association modes is realized, various interactive methods are designed to support a user to find abnormal modes of machine tool data, the association relation between abnormal changes of variables of the machine tool data and captured variables is analyzed, the abnormality can be quickly found and analyzed, and mutual influences, association and hidden correlation information among the machine tool parameters are fully mined.
Description
Technical Field
The invention relates to the field of detection of abnormal events of operation history data of a numerical control machine tool, in particular to a method and a system for correlation visual analysis of abnormal modes of operation data of the machine tool.
Background
The numerical control machine tool plays a vital role in the industrial field, can control the movement of the machine tool through a digital instruction, completes various complex part processing tasks, and is widely applied to the fields of military equipment, automobile manufacturing, mold processing, aerospace and the like. However, frequent malfunctions of machine tool equipment during operation can cause various problems including damage to the equipment itself, degradation of production efficiency, economic loss, and safety accidents.
With the advancement of technology, the time sequence symbolization technology provides an effective method for analysis tasks such as time sequence data classification, clustering, association rule mining and the like. Meanwhile, visual expression and analysis are widely applied in various industries, and the capability of better organizing, managing and understanding industry data is provided for each industry. The production running states of the machine tool can be recorded and detected in real time by various sensors, recorders and other devices, and a user can be helped to quickly and accurately determine fault points by analyzing and processing machine tool running data, and a corresponding solution is provided. The method not only can shorten maintenance downtime and improve production efficiency, but also can reduce maintenance cost. Therefore, the method has important significance in effectively monitoring, analyzing and maintaining the machine tool operation data.
Although some machine learning-based models can analyze and mine multidimensional time series data, the analysis and mining are excellent in detecting abnormal points or ranges, the analysis of association relation between abnormal changes of variables and capturing of the variables is lacking only by identifying abnormal points or ranges in the data and providing some statistical information to describe the abnormalities, and information provided in explaining abnormal reasons is very limited and cannot be rapidly and accurately analyzed.
Disclosure of Invention
Aiming at the problems in the field, the invention provides a visual analysis method and a visual analysis system for association of abnormal modes of machine tool operation data, which can solve the technical problems that the lack of analysis of association relation between abnormal changes of variables and captured variables, the information provided in the aspect of explanation of abnormal reasons is very limited, and the abnormal reasons cannot be rapidly and accurately analyzed.
In order to solve the technical problems, the invention discloses a machine tool operation data abnormal mode association visual analysis method and a system, wherein the method comprises the following steps:
respectively extracting symbol characteristics from a sliding window and a time point of original machine tool operation data, and performing multi-parameter type coding to obtain a multi-parameter coding matrix sequence;
extracting multi-parameter class code combinations of each column as data samples according to the multi-parameter code matrix sequence, performing cluster analysis and association rule mining, extracting a machine tool operation mode according to a cluster result obtained by the cluster analysis, and extracting a parameter state association mode according to an association rule set obtained by the association rule mining;
an interactive visual analysis system is constructed, the abnormal mode of machine tool data is found by supporting a user through designing a plurality of interactive methods, visual analysis mapping of multi-parameter type codes, machine tool running modes and parameter state association modes is realized, and the association relation between abnormal changes of variables of the machine tool data and captured variables is analyzed.
Preferably, the extracting the symbol features from the sliding window and the time point of the original machine tool operation data respectively includes the following steps:
the method comprises the steps of performing z-score standardization on a plurality of subsequences of original machine tool operation data, and performing standardization processing on each subsequence to obtain a subsequence with a mean value of 0 and a standard deviation of 1; subsequence s= { t 1 ,t 2 ,t 3 ,t 4 ...,t n Normalized sequence s is:
wherein μ is the average value among all the elements of S, σ is the standard deviation of the sequence;
the normalized multidimensional time series matrix C is obtained as follows:
wherein n is the length of the time sequence, and m is the number of machine tool parameters;
dividing the multidimensional time sequence matrix C according to a fixed sliding window dividing method, wherein w is the length of the sliding window, and obtaining a coding matrix sequence D as follows:
wherein n-w+1 is the number of sliding windows, d ij Is a sliding window average; s is(s) ij Normalized data for the jth time point of the ith parameter;
according to the breakpoint table, for the sliding window average value d ij The symbols of (2) are encoded to obtain an encoding matrix E as follows:
wherein e ij The sign code is the average value.
Preferably, the normalizing the plurality of sub-sequences of the original machine tool operation data using the z-score normalizes each sub-sequence, comprising the steps of:
encoding a plurality of parameter categories at each time point to obtain an encoding matrix F as follows:
wherein m represents a machine tool parameter, n represents a time series length, f ij Coding symbols for a point in time; for the coding of the time-point multiparameter class, the coding is represented by letters representing the division of different regions and numbers representing different parameters, i.e. the coding combination { f of the first column 11 ,f 21 ,f 31 ,f 41 The first parameter is in the d state at the first second = { d1, c2, b3, a4 }; the second and the third parameters are respectively in the states of c and b; the fourth parameter is in the a state.
Preferably, the extraction algorithm for extracting the operation mode of the machine tool comprises the following steps:
input: original machine tool operation data;
and (3) outputting: the type of machine tool operation mode;
step1: performing z-score standardization processing on the original machine tool operation data;
step2: dividing the original data according to a fixed sliding window dividing method;
step3: calculating the average value of sliding window data, coding by letters corresponding to the divided areas, and constructing a coding matrix sequence E, wherein each column represents a plurality of parameter states of one sliding window;
step4: extracting multi-parameter type code combinations of each column of the code matrix E, and clustering after de-duplicating the multi-parameter type code combinations;
step5: and dividing each sliding window coding combination into different clusters according to the clustering result so as to express a plurality of machine tool operation modes.
Preferably, the association rule extraction algorithm for association rule mining comprises the following steps:
input: original machine tool operation data;
and (3) outputting: a set of machine tool parameter state association rules;
step1: performing z-score standardization processing on the original machine tool operation data;
step2: encoding the data at each time point, the encoding being represented by letters representing different region divisions and numbers representing different parameters;
step3: constructing a coding matrix F, combining the coding results of each time point into a matrix according to columns to obtain a matrix containing a plurality of columns, wherein each column represents a plurality of parameter states of one time point;
step4: extracting multi-parameter class coding combinations of each column of the coding matrix F, applying an Apriori association rule algorithm to carry out mining, and setting a threshold value of the support degree and the confidence degree;
step5: and obtaining a set of association rules between machine tool parameter states according to the set support and confidence threshold.
Preferably, the setting the threshold of the support degree and the confidence degree includes the following steps:
the support degree calculation formula is:
wherein support (·) represents the degree of support;representing the relative probability that a and d occur simultaneously in all states;
the confidence coefficient calculation formula is:
wherein confidence (·) represents confidence;the probability of the d-state reappearance in the case where the a-state appears among all the states is represented.
Preferably, the extraction algorithm for extracting the parameter state association mode comprises the following steps:
extracting symbol characteristics from time points of original machine tool operation data and performing parameter class coding;
performing association rule mining on a plurality of parameter code combinations;
and expressing a plurality of parameter state association modes according to the association rule set.
Preferably, the system further comprises a visual analysis method for correlation of abnormal modes of machine tool operation data, and the system comprises:
the data acquisition module is used for respectively extracting symbol characteristics from a sliding window and a time point of the original machine tool operation data;
the data processing module is used for carrying out multi-parameter type coding according to the symbol characteristics extracted by the data acquisition module to obtain a multi-parameter coding matrix sequence; designing a plurality of interaction methods to support a user to find abnormal modes of machine tool data;
the data analysis module is used for extracting multi-parameter class code combinations of each column as data samples according to the multi-parameter code matrix sequences obtained by the data processing module, carrying out cluster analysis and association rule mining, extracting machine tool operation modes according to cluster results obtained by the cluster analysis, and extracting parameter state association modes according to association rule sets obtained by the association rule mining;
the data visual analysis module is used for constructing an interactive visual analysis system, analyzing the abnormal mode of the data processing module, realizing visual analysis mapping of multi-parameter type codes, machine tool running modes and parameter state association modes, and acquiring association relations between abnormal changes of variables of machine tool data and captured variables.
Preferably, the visual analysis system comprises a control panel view, a global view, a local view, a key index contrast view and a parameter state association view.
Compared with the prior art, the invention has the following beneficial effects:
1. the interactive visual analysis system constructed by the invention is embedded with the machine tool operation mode and parameter state association mode extraction algorithm, so that a user is supported to flexibly explore the data attribute change, and methods such as abnormal positioning, sensitive parameter analysis, state similarity search, parameter state association analysis and the like are provided for operation data monitoring, potential information in production process data is intuitively and efficiently extracted, and the user can carry out comprehensive interactive analysis on fault data by using the method designed by the invention.
2. The method has the advantages that through carrying out parameter category coding on machine tool operation data from two aspects of a sliding window and a time point and extracting symbol characteristics, according to a designed multiple interaction method, a user can be supported to find an abnormal mode of the machine tool data, and through analyzing abnormal changes of all variables of the machine tool data and capturing association relations among all variables, which parameters cause abnormality can be quickly found and analyzed.
3. By combining a visual analysis system of the mechineVis, a case study is carried out on a machine tool data set through visual analysis mapping and multi-view interaction linkage, and the visual analysis system of the invention is proved to be capable of easily finding and analyzing anomalies and fully excavating mutual influence, association and hidden correlation information among machine tool parameters.
Drawings
FIG. 1 is an overall schematic diagram of a visual analysis system mechineVis of the present invention;
FIG. 2 is a schematic diagram of a sliding window multi-parameter class encoding according to the present invention;
FIG. 3 is a schematic diagram of a multi-parameter code combination clustering method according to the present invention;
FIG. 4 is a schematic diagram of encoding a time-point multi-parameter class according to the present invention;
FIG. 5 is a schematic diagram of the time-point multi-parameter coding combination association rule mining according to the present invention;
FIG. 6 is a diagram of a visual analysis system of the mechineVis of the present invention;
FIG. 7 is a graph of analysis results of a global view of an embodiment portion of the present invention;
FIG. 8 is a graph of analysis results of key indicator views of an embodiment of the present invention;
FIG. 9 is a graph of analysis results of a partial view of an embodiment portion of the present invention;
FIG. 10 is a graph of analysis results of a parameter status association view of an embodiment portion of the present invention;
FIG. 11 is a graph showing analysis results of abnormal regions in the embodiment of the present invention;
fig. 12 is a graph showing the analysis result of the abnormal pattern in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to fig. 1 to 12 in the embodiments of the present invention. It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
In recent years, various feature extraction techniques including symbolization techniques have been developed in time series mining, in which a symbol aggregation approximation method is used in a large number, the method is to segment a time series based on a segment aggregation approximation method first, and use the mean value of each segment to represent the data of the original series, quantize the data of the original series into a plurality of regions according to probability using standard normal distribution, symbolize the values of different regions, and map the mean value onto the symbol. For visual analysis of anomaly detection, typically automated anomaly detection methods ignore subtle, ambiguous, uncertain anomalies, yet visual analysis improves the ability to analyze data through visual expression and visual interface of the data. Feature information is extracted through a time sequence symbolization technology, and an effective method and tool are provided for tasks such as classification, clustering, association rule mining, similarity measurement and the like of time sequence data. Meanwhile, visual expression and analysis are widely applied to various industries, and the capability of better organizing, managing and understanding industry data is provided for each industry. In the field of machine tool fault diagnosis, research work on visual analysis of abnormality detection of machine tools is not uncommon.
The visual analysis system mechineVis provided by the invention realizes visual analysis mapping of multi-parameter type codes, machine tool operation modes and parameter state association modes, designs a plurality of interaction methods to support users to find abnormal modes of machine tool data, analyzes state changes of all parameters of the machine tool data and captures association relations among all parameters, helps users to understand the machine tool operation data and analyze abnormal reasons, quickly positions problems, reduces downtime and improves production efficiency.
The invention provides a machine tool operation data abnormal mode association visual analysis method, which is realized through the following steps:
firstly, mapping multi-parameter class codes of a sliding window;
respectively extracting symbol characteristics from a sliding window and a time point of machine tool operation data and performing parameter category coding;
multiple sub-sequences of raw operational data of the motor bearing were normalized using z-score normalization, each sub-sequence normalized to a sub-sequence with a mean of 0 and standard deviation of 1. So for the subsequence s= { t 1 ,t 2 ,t 3 ,t 4 ...,t n Normalized sequence s is:
wherein μ is the average value of all elements in S, σ is the standard deviation of the sequence, and the normalized multidimensional time sequence matrix C is obtained and expressed as:
wherein n is the time sequence length, and m is the number of machine tool parameters.
Dividing the multidimensional time sequence matrix C according to a fixed sliding window dividing method, wherein w is the length of the sliding window, and a coding matrix sequence D can be obtained and expressed as follows:
wherein n-w+1 is the number of sliding windows, d ij Is a sliding window average; s is(s) ij Normalized data for the jth time point of the ith parameter;
sliding window average d ij The symbol codes of (2) are coded and mapped according to the discontinuity point table, and as shown in table 1, a coding matrix E is obtained as follows:
wherein e ij The sign code is the average value.
Table 1 breakpoint table
As can be seen from table 1, when a=4, there are three break points of-0.67, 0 and 0.67, respectively, which are divided into 4 areas. Wherein a portion having a value less than-0.67 is denoted by letter a, a portion having a value between-0.67 and 0 is denoted by letter b, a portion having a value between 0 and 0.67 is denoted by letter c, and a portion having a value greater than 0.67 is denoted by letter d. Wherein the sliding window multi-parameter class coding is shown in fig. 2.
Secondly, clustering a plurality of parameter coding combinations;
for coded matrix sequencesThe multi-parameter class code combination of each column is extracted as a data sample for cluster analysis, and as shown in fig. 3, the machine tool operation mode extraction algorithm comprises the following steps:
input: raw machine tool operation data
Output: multiple modes of operation of a machine tool
Step1: z-score normalization of raw machine tool operating data
Step2: dividing the original data according to a fixed sliding window dividing method
Step3: calculating average value of sliding window data, coding by letters of corresponding divided regions, constructing coding matrix sequence E, each column representing multiple parameter states of a sliding window
Step4: extracting multi-parameter type code combinations of each column of the code matrix E, de-duplicating the multi-parameter type code combinations, and clustering
Step5: and dividing each sliding window into different clusters according to the clustering result so as to express a plurality of machine tool operation modes.
The sample set is divided into a plurality of clusters through clustering, so that the data in the same cluster are high in similarity, and the difference between different clusters is large. Clustering is usually to calculate the Euclidean distance from each sample point to the clustering center to measure the similarity, but the distance cannot be calculated through numerical values by the letters after the sliding window data are coded, and clustering is performed by giving a distance table among the letters, as shown in Table 2, and looking up the distance table among the letters.
Table 2 distance meter
Thirdly, expressing a plurality of machine tool operation modes according to the clustering result;
extracting symbol characteristics from time points of machine tool operation data and performing parameter class coding; mapping the multi-parameter category codes of the time points; when the machine tool operation data is mined, in order to fully mine the relation among the parameters, after the machine tool operation data is standardized, a plurality of parameter categories at each time point are encoded, and an encoding matrix F is obtained as follows:
wherein m represents a machine tool parameter, n represents a time series length, f ij Encoding the symbols for the time points.
For point-in-time multiparameter classesOther codes are represented by letters representing different region divisions and numbers representing different parameters. The code combination { f as the first column 11 ,f 21 ,f 31 ,f 41 The first parameter is in the d state at the first second = { d1, c2, b3, a4 }; the second and the third parameters are respectively in the states of c and b; the fourth parameter is in the a state; the resulting time-point multi-parameter class code is shown in fig. 4.
Performing association rule mining on a plurality of parameter code combinations;
and generating a state association rule set among parameters by applying an Apriori association rule algorithm aiming at the time point multi-parameter class coding combination of each column in the coding matrix sequence F. The method consists essentially of two parts, namely a support degree and a confidence degree.
The support is the relative probability of the different states a and d of two different parameters occurring simultaneously, namely:
wherein support (·) represents the degree of support;
confidence represents the probability of the d-state reappearance in the case of the a-state appearing among all states, namely:
wherein confidence (·) represents confidence;
the user can autonomously set the threshold value of the support degree and the confidence degree according to the actual needs and the data characteristics so as to screen out the association rule with certain meaning and credibility.
As shown in fig. 5, for n parameters in the machine tool operation data, m letters are used to encode to represent the status thereof. Thus, each parameter has m states, for a total of m×n possible states; and calculating the support degree and the confidence degree between different parameter states, and screening out the association rule meeting the threshold requirement, thereby finding out the association relation between the parameter states.
The machine tool parameter state association rule extraction algorithm comprises the following steps:
input raw machine tool operation data
Output machine tool parameter state association rule set
Step1: z-score normalization of raw machine tool operating data
Step2: the data at each point in time is encoded, the encoding being represented by letters representing the division of different regions and numbers representing different parameters
Step3: constructing a coding matrix F, combining the coding results of each time point into a matrix according to columns, thus obtaining a matrix containing a plurality of columns, wherein each column represents a plurality of parameter states of one time point
Step4: extracting multi-parameter class code combinations of each column of the code matrix F, applying an Apriori association rule algorithm to carry out mining, and setting a threshold value of support and confidence
Step5: and obtaining a set of association rules between machine tool parameter states according to the set support degree and confidence degree threshold.
According to the invention, a plurality of parameter state association modes are expressed according to an association rule set, detailed design and interaction of each view in a visual analysis system mechineVis system are realized, and a plurality of mode extraction algorithms are mapped and embedded through visual analysis of multi-parameter class codes, so that an interactive exploration tool is provided for researchers. The mechineVis system mainly contains five key views, as shown in FIG. 6, and specifically includes: (a) a control panel view, (b) a local view, (c) a key index contrast view, (d) a parameter status association view, and (e) a global view.
As shown in fig. 6 (a), the control panel view includes five options from top to bottom: data, number of clusters, sliding window length, support and confidence are selected. Selecting data for determining machine tool operation data to be analyzed, wherein the selection of the number of clusters and the length of a sliding window is related to the extraction of machine tool operation modes, the number of clusters represents the number of the machine tool operation modes, and the length of the sliding window represents the time sequence span of the machine tool operation data segmented according to the sliding window; the magnitude of the support and the confidence are set to be related to the extraction of the association mode of the parameter states of the machine tool, and the higher the support and the confidence, the stronger the association between the representing parameter states.
As shown in fig. 6 (e), the global view shows the operation data of the machine tool in the form of a line graph, and the hatched area of the line graph represents the magnitude and duration of the fluctuation of the data. The background color of the mouse positioning represents the running state of the current machine tool, when the user positions the mouse at a certain moment, the running state of the bearing is represented by different colors, and if the number of the symbol letters is 4, the running state of the bearing is represented by 4 colors. When a user clicks a state at a certain moment, the global view can search a time segment identical to the current machine tool running state, and the similar search implementation method can be used for matching the same running state by judging whether the coding combinations are identical.
According to the clustering result of the sliding window multi-parameter coding combination, different clusters can represent different running modes in the machine tool work, and different clusters are mapped by different colors. In fig. 6 (e), the rectangle at the bottom of the view shows the machine tool operating mode change, which can provide an intuitive description and general overview of the bearing operating state. If the color of the rectangular bar changes, it represents that the machine tool operation mode has changed. In addition, a time brush is provided in the global view, and the display range of the local view is adjusted by means of a sliding block, mouse dragging and the like.
As shown in fig. 6 (b), by interpreting the partial view, the user can more carefully observe a specific area where the abnormality is located and further analyze the cause of the abnormality. The left half part of the partial view shows the detailed change condition of each parameter data in a partial time period, each rectangle represents the current time state, and the detailed information of each parameter operation data change can be mined according to the color change position of the rectangle.
The color rectangle bar at the bottom of the partial view represents the running mode of the bearing in the selected time period, different running modes are distinguished by colors, and a user can analyze running data of each parameter in the abnormal mode more conveniently.
Also in fig. 6 (b), a comparative analysis of the state at a certain time and the previous state is provided. When the mouse is placed at a certain time, as shown in the left part of the diagram (b) in fig. 6, two small rectangles will display the state of the previous second and the running state of the previous two seconds, and the large rectangle represents the current running state of the mouse positioning. If the size rectangles differ in color, the state representing the current parameter changes relative to the first two seconds. As shown in the thermodynamic diagram in the right half of fig. 6 (b), which represents the correlation between different parameters over a selected period of time, different colors in the thermodynamic diagram represent different correlations, e.g., red represents a higher positive correlation and purple represents a higher negative correlation. In addition, a correlation index screening function is also provided in the partial view.
As shown in the (d) diagram in fig. 6, the association view of parameter states is shown, in which the association between each parameter state in the machine tool operation data can be fully mined, and after association rule mining is performed on the multi-parameter coding combination, the screened association rule is visually analyzed and displayed.
Each circle represents a state, and circles of different colors represent different states, so that a user can quickly distinguish between different parameter states. The line between circles represents an association rule. The line color shade represents the support level and the line width represents the confidence level. Wherein, the deeper the line, the higher the support, and the wider the line width, the higher the confidence. After clicking one state, the user can analyze the global view and find out the association relation among different states of each parameter and the occurrence time of state association.
As shown in fig. 6 (c), a key index comparison view is shown, in which the comparison situation of the characteristic values of each parameter of the machine tool in different operation modes is shown by using a radar chart, and the characteristic values of the comparison include an average value, a variance and a standard deviation. By comparing the characteristic values of the parameters in different modes, the state change condition of the parameters of the machine tool in different operation modes can be found, so that the operation state of the machine tool can be better estimated.
The invention also provides a system of the machine tool operation data abnormal mode association visual analysis method, which comprises the following steps:
the data acquisition module is used for respectively extracting symbol characteristics from a sliding window and a time point of the original machine tool operation data;
the data processing module is used for carrying out multi-parameter type coding according to the symbol characteristics extracted by the data acquisition module to obtain a multi-parameter coding matrix sequence; designing a plurality of interaction methods to support a user to find abnormal modes of machine tool data;
the data analysis module is used for extracting multi-parameter class code combinations of each column as data samples according to the multi-parameter code matrix sequences obtained by the data processing module, carrying out cluster analysis and association rule mining, extracting machine tool operation modes according to cluster results obtained by the cluster analysis, and extracting parameter state association modes according to association rule sets obtained by the association rule mining;
the data visual analysis module is used for constructing an interactive visual analysis system, analyzing the abnormal mode of the data processing module, realizing visual analysis mapping of multi-parameter type codes, machine tool running modes and parameter state association modes, and acquiring association relations between abnormal changes of variables of machine tool data and captured variables.
The visual analysis system comprises a mechineVis system, is used for visual analysis mapping of multi-parameter class codes obtained by the data analysis module, and is embedded into a multi-mode extraction algorithm of the data processing module.
Wherein, the mechineVis system includes: control panel view, global view, local view, key index contrast view, parameter state association view.
The invention extracts the symbol characteristics from the sliding window of the machine tool operation data and the time point and combines the parameters, thereby realizing the coding of the machine tool operation state. And clustering the sliding window multi-parameter coding combinations to identify different machine tool operation modes, performing association rule mining on the time point multi-parameter coding combinations, calculating the support and confidence among the parameter states, and further analyzing the association relation in the data. Finally, a case study is carried out on the machine tool data set through visual analysis mapping and multi-view interactive linkage by combining a visual analysis system mechineVis, so that the visual analysis system can easily find and analyze anomalies, and can fully mine the mutual influence, association and hidden correlation information among machine tool parameters.
Examples
The present invention uses a published dataset SKAB, which is a multivariate time series collected from sensors mounted on the motor test stand, containing eight parameters, acceleromer 1RMS (vibration acceleration), acceleromer 2RMS (vibration acceleration), current, pressure, temperature of the circuit fluid, voltage, volumeFlowRate RMS (circulation flow rate of the circuit fluid), respectively. The data recording time is 2020-02-0817:27:19 to 2020-02-0817:47:18, each second of data, is 1144 in length. First, the data needs to be subjected to z-score normalization, the number of letters is set to 4 in the control panel view, the length of the sliding window is set to 30, and the number of clusters is set to 3. While the support was set to 0.25 and the confidence was set to 0.85.
As can be seen from the rectangle at the bottom of the global view, the machine tool operating mode is divided into three modes, mode 1, mode 2 and mode 3, mapped with different colors, and at this point in time 585 the machine tool operating mode is found to have changed. In the region of pattern 2, the shadow areas of the parameters A1RMS (abbreviation for Accelerometer 1 RM), A2RMS (abbreviation for Accelerometer 2 RM), VRMS (abbreviation for VolumeFlowRate RMS) are clearly observed to be large, indicating that during this time period these parameters fluctuate widely and with a long duration, anomalies are likely to occur. Therefore, special attention needs to be paid to the distribution of these parameter states during this period. Dragging the time brush in the thermodynamic diagram of the bottom right portion of the global view, locating the time range in the region of pattern 2, it can be seen by looking at the local view that the red areas of A1RMS and A2RMS are widely covered, the bluish areas of VRMS also occupy most of the region, indicating that during this time, the data of A1RMS and A2RMS are relatively high, while the majority of the data of VRMS are relatively low, and it can be better determined that anomalies have occurred in the time periods of the regions of pattern 2 for A1RMS, A2RMS, VRMS. By looking at the correlation thermodynamic diagram of the left part of the partial view, it was found that there was a positive correlation between A1RMS and A2RMS during this time, and that there was also a higher negative correlation between A1RMS and A2RMS and VRMS.
As shown in fig. 8, in the key index view, it is found by radar chart comparison that: in the three modes of operation, the eigenvalues of Current, voltage and Pressure are relatively close, indicating that Current, voltage and Pressure are in steady state throughout operation. However, A1RMS, A2RMS, VRMS, temperature and thermo couple have a large gap in the multiple eigenvalues in different modes of operation. The judgment of A1RMS, A2RMS and VRMS in conjunction with the front view of fig. 7 may be because the machine tool operating state is not stable enough or is affected by abnormal factors. And the Temperature and the thermo-couple show overall descending trend, so that a larger difference between characteristic values in different modes is caused.
As shown in fig. 9, the three anomaly sequences of A1RMS, A2RMS and VRMS are analyzed in detail in the partial view, and it is found that before the machine tool is abnormal, the color arrangement combination of the rectangular bars of A1RMS and A2RMS is changed greatly in the early stage, and the state of A2RMS and A1RMS is changed greatly from 497 seconds and 500 seconds respectively in combination with the change condition of the rectangle of the size and the mouse positioning. Also after the abnormality of the machine tool, it was found that VRMS began to gradually decrease from 596 seconds and the state was changed considerably from before. It can thus be inferred that: the continuous rise in A1RMS and A2RMS may be responsible for machine tool anomalies, such that VRMS anomalies are caused.
As shown in fig. 10, in the parameter state association view, A1RMS (d), A2RMS (d), VRMS (a) are found to be associated with each other, and after one of the states is clicked with a mouse, the position where each state appears can be displayed in the global view. As shown in fig. 11, the areas where the association states of A1RMS (d), A2RMS (d), and VRMS (a) appear are found to be within the abnormal interval, i.e., it can be inferred that A1RMS, A2RMS, and VRMS will generally produce abnormalities together, and VRMS is lower when A1RMS and A2RMS are higher.
Where A1RMS (a) and A2RMS (a) are also interrelated, indicating that in mode 1 and mode 3, when A1RMS is lower, A2RMS is also lower. It was also found that Temperature (d) and Thermocouple (d) point to A1RMS (a) simultaneously, indicating that in mode 1, A1RMS is typically lower when Temperature is higher and when thermo couple is higher.
As shown in FIG. 12, when clicking on the background for 769 seconds in the light red region of the global view, it can be seen that this operational state is also present at times 775, 788, 865, 887, 896, 903, 930, 952, and the same abnormal pattern occurrence time can be observed, indicating that there are numerous similar operational states in the interval of [769-952 ].
The invention respectively extracts symbol characteristics from a sliding window and a time point of the original machine tool operation data and combines parameters to realize the coding of machine tool parameter types; clustering the sliding window multi-parameter coding combinations, identifying different machine tool operation modes, performing association rule mining on the time point multi-parameter coding combinations, calculating the support degree and the confidence degree among parameter states, and further analyzing the association relation in the data; by combining an interactive visual analysis system mechineVis, and through visual analysis mapping and multi-view interactive linkage, case study is conducted on a machine tool data set, and the visual analysis system is proved to be capable of easily finding and analyzing which parameters cause abnormality and fully mining mutual influence, association and hidden correlation information among machine tool parameters.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.
In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. All documents mentioned in this specification are incorporated by reference for the purpose of disclosing and describing the methodologies associated with the documents. In case of conflict with any incorporated document, the present specification will control.
Claims (9)
1. The machine tool operation data abnormal mode association visual analysis method is characterized by comprising the following steps of:
respectively extracting symbol characteristics from a sliding window and a time point of original machine tool operation data, and performing multi-parameter type coding to obtain a multi-parameter coding matrix sequence;
extracting multi-parameter class code combinations of each column as data samples according to the multi-parameter code matrix sequence, performing cluster analysis and association rule mining, extracting a machine tool operation mode according to a cluster result obtained by the cluster analysis, and extracting a parameter state association mode according to an association rule set obtained by the association rule mining;
an interactive visual analysis system is constructed, the abnormal mode of machine tool data is found by supporting a user through designing a plurality of interactive methods, visual analysis mapping of multi-parameter type codes, machine tool running modes and parameter state association modes is realized, and the association relation between abnormal changes of variables of the machine tool data and captured variables is analyzed.
2. A machine tool operation data abnormal pattern correlation visual analysis method according to claim 1, wherein said extracting symbol features from a sliding window and a time point of original machine tool operation data, respectively, comprises the steps of:
the method comprises the steps of performing z-score standardization on a plurality of subsequences of original machine tool operation data, and performing standardization processing on each subsequence to obtain a subsequence with a mean value of 0 and a standard deviation of 1; subsequence s= { t 1 ,t 2 ,t 3 ,t 4 ...,t n Normalized sequence s is:
wherein μ is the average value among all the elements of S, σ is the standard deviation of the sequence;
the normalized multidimensional time series matrix C is obtained as follows:
wherein n is the length of the time sequence, and m is the number of machine tool parameters;
dividing the multidimensional time sequence matrix C according to a fixed sliding window dividing method, wherein w is the length of the sliding window, and obtaining a coding matrix sequence D as follows:
wherein n-w+1 is the number of sliding windows, d ij Is a sliding window average; s is(s) ij Normalized data for the jth time point of the ith parameter;
according to the breakpoint table, for the sliding window average value d ij The symbols of (2) are encoded to obtain an encoding matrix E as follows:
wherein e ij The sign code is the average value.
3. A machine tool operation data abnormal pattern correlation visual analysis method according to claim 2, wherein said normalizing each sub-sequence of said plurality of sub-sequences of said original machine tool operation data using z-score normalization comprises the steps of:
encoding a plurality of parameter categories at each time point to obtain an encoding matrix F as follows:
wherein m represents a machine tool parameter, n represents a time series length, f ij Coding symbols for a point in time; for the coding of the time-point multiparameter class, the coding is represented by letters representing the division of different regions and numbers representing different parameters, i.e. the coding combination { f of the first column 11 ,f 21 ,f 31 ,f 41 The first parameter is in the d state at the first second = { d1, c2, b3, a4 }; the second and the third parameters are respectively in the states of c and b; the fourth parameter is in the a state.
4. A machine tool operation data abnormal pattern correlation visual analysis method according to claim 2, wherein said extraction algorithm for extracting machine tool operation patterns comprises the steps of:
input: original machine tool operation data;
and (3) outputting: the type of machine tool operation mode;
step1: performing z-score standardization processing on the original machine tool operation data;
step2: dividing the original data according to a fixed sliding window dividing method;
step3: calculating the average value of sliding window data, coding by letters corresponding to the divided areas, and constructing a coding matrix sequence E, wherein each column represents a plurality of parameter states of one sliding window;
step4: extracting multi-parameter type code combinations of each column of the code matrix E, and clustering after de-duplicating the multi-parameter type code combinations;
step5: and dividing each sliding window coding combination into different clusters according to the clustering result so as to express a plurality of machine tool operation modes.
5. A machine tool operation data anomaly pattern association visual analysis method according to claim 3, wherein the association rule extraction algorithm of association rule mining comprises the following steps:
input: original machine tool operation data;
and (3) outputting: a set of machine tool parameter state association rules;
step1: performing z-score standardization processing on the original machine tool operation data;
step2: encoding the data at each time point, the encoding being represented by letters representing different region divisions and numbers representing different parameters;
step3: constructing a coding matrix F, combining the coding results of each time point into a matrix according to columns to obtain a matrix containing a plurality of columns, wherein each column represents a plurality of parameter states of one time point;
step4: extracting multi-parameter class coding combinations of each column of the coding matrix F, applying an Apriori association rule algorithm to carry out mining, and setting a threshold value of the support degree and the confidence degree;
step5: and obtaining a set of association rules between machine tool parameter states according to the set support and confidence threshold.
6. The visual analysis method for abnormal pattern association of machine tool operation data according to claim 5, wherein the setting of the threshold of the support degree and the confidence degree comprises the steps of:
the support degree calculation formula is:
wherein support (·) represents the degree of support;representing the relative probability that a and d occur simultaneously in all states;
the confidence coefficient calculation formula is:
wherein confidence (·) represents confidence;the probability of the d-state reappearance in the case where the a-state appears among all the states is represented.
7. A machine tool operation data anomaly pattern correlation visual analysis method according to claim 5, wherein the extraction algorithm for extracting parameter state correlation patterns comprises the steps of:
extracting symbol characteristics from time points of original machine tool operation data and performing parameter class coding;
performing association rule mining on a plurality of parameter code combinations;
and expressing a plurality of parameter state association modes according to the association rule set.
8. A system for machine tool operation data anomaly pattern correlation visual analysis method, comprising:
the data acquisition module is used for respectively extracting symbol characteristics from a sliding window and a time point of the original machine tool operation data;
the data processing module is used for carrying out multi-parameter type coding according to the symbol characteristics extracted by the data acquisition module to obtain a multi-parameter coding matrix sequence; designing a plurality of interaction methods supports a user to find abnormal modes of machine tool data:
the data analysis module is used for extracting multi-parameter class code combinations of each column as data samples according to the multi-parameter code matrix sequences obtained by the data processing module, carrying out cluster analysis and association rule mining, extracting machine tool operation modes according to cluster results obtained by the cluster analysis, and extracting parameter state association modes according to association rule sets obtained by the association rule mining;
the data visual analysis module is used for constructing an interactive visual analysis system, analyzing the abnormal mode of the data processing module, realizing visual analysis mapping of multi-parameter type codes, machine tool running modes and parameter state association modes, and acquiring association relations between abnormal changes of variables of machine tool data and captured variables.
9. The system of claim 8, wherein the visual analysis system comprises a control panel view, a global view, a local view, a key index contrast view, and a parameter status association view.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310585965.0A CN116628611A (en) | 2023-05-23 | 2023-05-23 | Visual analysis method and system for association of abnormal modes of machine tool operation data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310585965.0A CN116628611A (en) | 2023-05-23 | 2023-05-23 | Visual analysis method and system for association of abnormal modes of machine tool operation data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116628611A true CN116628611A (en) | 2023-08-22 |
Family
ID=87602020
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310585965.0A Pending CN116628611A (en) | 2023-05-23 | 2023-05-23 | Visual analysis method and system for association of abnormal modes of machine tool operation data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116628611A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117102950A (en) * | 2023-10-17 | 2023-11-24 | 上海诺倬力机电科技有限公司 | Fault analysis method, device, electronic equipment and computer readable storage medium |
-
2023
- 2023-05-23 CN CN202310585965.0A patent/CN116628611A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117102950A (en) * | 2023-10-17 | 2023-11-24 | 上海诺倬力机电科技有限公司 | Fault analysis method, device, electronic equipment and computer readable storage medium |
| CN117102950B (en) * | 2023-10-17 | 2023-12-22 | 上海诺倬力机电科技有限公司 | Fault analysis method, device, electronic equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hsu et al. | Multiple time-series convolutional neural network for fault detection and diagnosis and empirical study in semiconductor manufacturing | |
| Ma et al. | Supervised anomaly detection in uncertain pseudoperiodic data streams | |
| Liu et al. | Noise removal in the presence of significant anomalies for industrial IoT sensor data in manufacturing | |
| Hachicha et al. | A survey of control-chart pattern-recognition literature (1991–2010) based on a new conceptual classification scheme | |
| JP5331774B2 (en) | Equipment state monitoring method and apparatus, and equipment state monitoring program | |
| US8682824B2 (en) | Method and device for monitoring the state of a facility | |
| JP5740459B2 (en) | Equipment status monitoring method | |
| JP5301310B2 (en) | Anomaly detection method and anomaly detection system | |
| WO2010041355A1 (en) | Error detection method and system | |
| WO2013011745A1 (en) | Facility state monitoring method and device therefor | |
| KR20150018642A (en) | Generalized pattern recognition for fault diagnosis in machine condition monitoring | |
| JP2013218725A (en) | Abnormality detecting method and abnormality detecting system | |
| Yang et al. | Refined composite multivariate multiscale symbolic dynamic entropy and its application to fault diagnosis of rotating machine | |
| CN116628611A (en) | Visual analysis method and system for association of abnormal modes of machine tool operation data | |
| EP3333757B1 (en) | Predictive anomaly detection | |
| JP5498540B2 (en) | Anomaly detection method and system | |
| CN108647737A (en) | A kind of auto-adaptive time sequence variation detection method and device based on cluster | |
| CN116756595A (en) | Conductive slip ring fault data acquisition and monitoring method | |
| CN117290802B (en) | Host power supply operation monitoring method based on data processing | |
| Baek et al. | Abrupt variance and discernibility analyses of multi-sensor signals for fault pattern extraction | |
| CN115524002A (en) | Running state early warning method and system for power plant rotating equipment and storage medium | |
| Alsallakh et al. | A Visual Analytics Approach to Segmenting and Labeling Multivariate Time Series Data. | |
| CN116149896B (en) | Time sequence data abnormality detection method, storage medium and electronic device | |
| Golosinski et al. | Data mining vims data for information on truck condition | |
| CN116431873A (en) | Visual analysis system for equipment operation data |
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 |