CN112171376A - Machine tool workpiece real-time statistical method based on current signal segmentation - Google Patents
Machine tool workpiece real-time statistical method based on current signal segmentation Download PDFInfo
- Publication number
- CN112171376A CN112171376A CN202010855703.8A CN202010855703A CN112171376A CN 112171376 A CN112171376 A CN 112171376A CN 202010855703 A CN202010855703 A CN 202010855703A CN 112171376 A CN112171376 A CN 112171376A
- Authority
- CN
- China
- Prior art keywords
- data
- processing
- model
- training
- segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
- B23Q17/00—Arrangements for observing, indicating or measuring on machine tools
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
- G01R19/0092—Arrangements for measuring currents or voltages or for indicating presence or sign thereof measuring current only
Abstract
The invention provides a real-time statistical method for machine tool workpieces based on current signal segmentation, which comprises two stages: a segmentation model training stage and a workpiece on-line segmentation statistical stage. Acquiring and processing data in a segmentation model training stage, firstly acquiring historical current signals and historical work order information of a machine tool, and cleaning and normalizing data to obtain a training sample; model training, namely training a sample by using a GaussianHMM model, and storing the trained model as a predictor; the on-line segmentation and statistics stage of the workpiece: and (3) segmenting the current signal by adopting a GaussianHMM model, extracting a single workpiece processing signal segment, and realizing online real-time statistics of the workpiece. The workpiece segmentation algorithm based on the model can realize accurate and real-time online statistics of workpieces under the condition of not establishing a workpiece processing signal template library in advance.
Description
[ technical field ] A method for producing a semiconductor device
The invention relates to the technical field of industrial machine tool machining, in particular to a real-time statistical method for machine tool workpieces based on current signal segmentation.
[ background of the invention ]
The traditional mode of machine tool workpiece processing progress information statistics is mainly through manual statistics, and especially under the condition that small-size processing factory often changes processing work piece type and quantity, manual statistics inefficiency just makes mistakes easily. The artificial statistics method has the following problems: firstly, after-work statistics is carried out, the real-time performance is poor, and the real-time progress condition of a processing field cannot be dynamically mastered; secondly, manual statistics is prone to errors.
Some new methods carry out statistics by collecting signals of equipment, and mainly solve the statistics of the progress of batch production of the same workpieces or large workpieces. The method for identifying and counting different types of workpieces based on the approximate matching algorithm of dynamic time warping is proposed in the literature (a method for automatically identifying and monitoring online processing workpieces based on dynamic time warping and power information; volume 47, No. 3 of the mechanical engineering report). In order to implement the method, firstly, the problem to be solved is how to extract a signal segment of a processing workpiece from a time series signal, wherein the starting and stopping of a motor through a fixed criterion are used as an extraction standard of the processing workpiece. However, in many industrial enterprises, the motor of the common machine tool is always in a starting state, and a machined workpiece cannot be simply divided by starting and stopping the machine; for a numerical control machine tool, a workpiece is also in a short-time starting and stopping state in the machining process, and the workpiece is cut into a plurality of parts by adopting the mode, so that the statistical error is caused. The second problem, in terms of data acquisition processing, is that the above method requires an acquisition frequency of 20Hz and an accuracy of 12 bits; the requirements on acquisition equipment and processing capacity are high, the acquisition cost is high, and the method is difficult to popularize in general industrial enterprises.
[ summary of the invention ]
The invention aims to solve the problems in the prior art, and provides a machine tool workpiece real-time statistical method based on current signal segmentation.
In order to achieve the purpose, the invention provides a real-time statistical method for machine tool workpieces based on current signal segmentation, which comprises the following steps:
a segmentation model training stage:
s11: collecting and processing data, namely collecting a historical current signal and historical work order information of a machine tool, and cleaning and normalizing the data to obtain a training sample;
s12: training a model, namely training a sample by using a GaussianHMM model, saving the trained model as a predictor, acquiring a segmentation point, and obtaining a segmented current signal; extracting statistical characteristics from the segmented current signals, setting a threshold value, and screening out workpiece processing signals;
the on-line segmentation and statistics stage of the workpiece:
s21: defining a cache region, and temporarily storing data in the cache region under the condition that no partition point exists after the same order data is transmitted into a partition model for prediction; predefining other variables;
s22: the current signal and the work order number information which are collected in real time are obtained in an interface mode, and the collection frequency is low, the highest frequency is 1S, the dimensionality is small, and therefore the requirement on storage configuration is not high;
s23: calling a pre-trained model, and transmitting cleaned data into model prediction; extracting break points according to different predicted states, wherein the break points are points of state change in a prediction result; acquiring a segment after current signal segmentation;
s24: sequentially traversing the segmentation segments, and extracting the statistical characteristics of each segment: the prediction state, the length and the mean value of each section are popped out from the queue in sequence, whether each section is a complete processing workpiece is judged based on the characteristic combination, and the judgment rule is a threshold value generated in the training stage;
s25: a change judgment and processing step of repeating steps S21 to S24 if no new order data is introduced; if a new order is introduced, some of the predefined variables of step S21 are set to null.
Preferably, in step S11, the data is collected by the gateway to collect a historical current signal and historical work order information of the machine tool, the frequency is 1S, the current signal includes a timestamp and a current value, and the historical work order information includes a work order number, a work order start time, and a work order end time.
Preferably, in step S11, before the data is cleaned, the two types of data are associated according to the time field, and the associated fields are: time stamp, current value, work order number, work order start time, and work order end time.
Preferably, the cleaning and normalization of the data in step S11 specifically includes the following steps:
s11.1, missing value processing, namely processing the timestamp into 1S (packet loss occurs in the acquisition report and needs to be processed), and completing the missing current value in a backward filling mode;
s11.2 processing abnormal values, wherein the unified processing that the current value is less than 0 is 0; unifications with quantiles greater than 90% are filled with quantiles of 90%; the current value after treatment is basically stabilized in the range of about 0-10A;
s11.3, normalization processing is carried out, wherein in order to accelerate the model tuning speed, normalization processing of a current value is required, and the normalization processing mode adopts min-max normalization;
s11.4, the data is deformed, and the shape of the input data is changed into a form of multiple rows and 1 column.
Preferably, step S12 specifically includes the following steps:
s12.1 input data: inputting the current value of the data after completion of the cleaning in step S11;
s12.2 model selection: training input data by adopting a GaussianHMM model;
s12.3, setting model parameters: the states of the data analysis equipment at the early stage are 3 types: stopping, waiting and processing, so the number of the hidden states is set to be 3, the iteration turn is set to be 100, and other parameters are defaulted;
s12.4 model results are returned: after training is finished, the model is stored locally, and meanwhile, the mean value and the standard deviation predicted as the processing state in the model are returned; and returning the prediction result of the input data as the training result of the model;
s12.5, training result processing and processed workpiece statistics: splicing input data and a training result (equipment state), and finding out all discontinuous points according to the training result, wherein the discontinuous points are defined as first points of result state change; the number of workpieces processed can be counted based on order number, order time (start and end time), equipment status and length.
Preferably, the collected signals are processed before the step S23, the modeling data is constructed, and the main tasks include filling missing values (collection inevitably causes packet loss), processing the instantaneous abnormal values of the current (using 90% upper quantile), and normalizing the current data.
Preferably, in step S23, according to whether there is a slot point in the prediction result, the following processes are respectively performed:
a1. if there is no break point, merging the prediction result with the data of the direct cache area, and storing the data in the cache area;
a2. if there is a break point, the prediction result is merged with the buffer area, and then the prediction state of the segment formed by each break point is traversed in sequence, the obtained segment is stored in a queue, and the data before the next break point is shifted out from the buffer area.
The method is based on the collected low-frequency time sequence current signal (the frequency is 1S) of the machine tool equipment motor, is related to the starting time and the ending time of a work order provided by software, adopts a GaussianHMM model to segment the current signal, extracts a single workpiece machining signal segment, and realizes the online real-time statistics of the workpiece. The accurate and real-time online statistics of the workpiece can be realized without establishing a workpiece processing signal template library in advance.
Compared with the prior art, the invention has the following beneficial effects:
the model has low requirement on input data: the data acquisition frequency is required to be only 1S once, so that the acquisition technical requirement is lower, and the cost for storing data is lower from the viewpoint of storing data.
The model realizes a workpiece segmentation algorithm based on the model, and can realize accurate and real-time online statistics of workpieces under the condition of not establishing a workpiece processing signal template library in advance.
The features and advantages of the present invention will be described in detail by embodiments in conjunction with the accompanying drawings.
[ description of the drawings ]
FIG. 1 is a block diagram of the overall process of model training and online use of the workpiece online statistical method of the present invention;
FIG. 2 is a detailed flow chart of each link of online statistical online usage of workpieces in the present invention.
[ detailed description ] embodiments
Aiming at the problem that a simple rule cannot accurately extract a machining signal section of a workpiece from an acquired time sequence signal, a machine learning model GaussianHMM model is adopted to extract a machining signal based on an acquired low-frequency current signal of a machine tool equipment motor and the start time and the end time of a work order provided by software, and the real-time online statistics of the workpiece divided into the work orders is realized without establishing a template library in advance.
In order to facilitate understanding of the technical scheme of the invention, a detailed description is given below by taking an example of a current signal of a real working environment acquired by a gateway additionally installed on a machine tool of a certain factory and data of start and end times of a work order acquired through software operation.
The specific embodiment is used for carrying out real-time online statistics on the number of the cut workpieces of a common cutting machine tool in a certain factory.
The whole implementation process is divided into two stages: a training phase and an online operation phase.
The specific process of the training phase is as follows:
step 1.1, data acquisition, namely respectively reading current information (including a timestamp and a current value) collected by a gateway in a certain time period and data (a work order number, a work order starting time and a work order ending time) of a work order recorded by software from a database.
Step 1.2 basic data preparation: and associating the two types of data according to the time field, wherein the associated field is as follows: time stamp, current value, work order number, work order start time, and work order end time.
Step 1.3, data cleaning: 1) missing value processing, namely processing the timestamp into 1S (packet loss occurs in the acquisition report and needs to be processed), and completing the missing current value in a backward filling mode; 2) abnormal value processing, wherein the unified processing that the current value is less than 0 is 0; unifications with quantiles greater than 90% are filled with quantiles of 90%; the current value after treatment is basically stabilized in the range of about 0-10A; 3) normalization processing, namely performing normalization processing on a current value in order to accelerate the model tuning speed, wherein the normalization processing mode is min-max normalization; 4) and transforming the data into a form of multiple rows and 1 column by the shape of the input data.
Step 1.4 model training:
inputting data: the current value of the data after cleaning in the step 1.3;
selecting a model: training input data by adopting a GaussianHMM model;
setting model parameters: the states of the data analysis equipment at the early stage are 3 types: stopping, waiting and processing, so the number of the hidden states is set to be 3, the iteration turn is set to be 100, and other parameters are defaulted;
and returning a model result: after training is finished, the model is stored locally, and meanwhile, the mean value and the standard deviation predicted as the processing state in the model are returned; and returning the prediction result of the input data as the training result of the model.
Step 1.5 training result processing and processed workpiece statistics
Splicing input data and a training result (equipment state), finding out all discontinuous points according to the training result, wherein the discontinuous points are defined as first points of result state change; counting the mean value and the length of each section according to the discontinuity points, wherein the mean value is larger than (max (standard deviation of the mean value of the machining state being-1.5 times), and 0.5) corresponding equipment state is work _ state (working state); according to business experience, the length requirement of a processing section is required to be more than 10; the number of workpieces processed can be counted based on the order number, the order time (start and end time), the equipment status and the length.
The specific flow of the online operation stage is as follows:
step 2.1 predefining variables
The basic parameters include: the method comprises the steps of (1) saving work _ state during training, counting _ num of an order counting variable, normal machining signal segment normal _ segment and abnormal machining signal segment normal _ segment;
defining a buffer area: after the same order data is transmitted and enters model prediction, if the prediction result has no dividing point, the data is temporarily stored in a cache region; the cache region stores the time stamp and the prediction result in a paired mode; considering the size of the cache area, if the length of the cache area is greater than 3600(1 hour) and the cache area is in a non-processing state, the front 5/6 in the cache area is discarded;
step 2.2 data acquisition
Transmitting current information (including a timestamp and a current value) collected by the gateway in the period of time every 30S, and transmitting data (a work order number, work order starting time and work order ending time) of a work order recorded by software if a new work order exists;
step 2.3 data cleaning
Cleaning the reserved 90% quantile, maximum value and minimum value according to the training data, and cleaning the data according to the data cleaning process in the step 1.3 of the flow of the training stage;
step 2.4 calling model prediction
Calling a pre-trained model, and transmitting the data cleaned in the step 2.3 into the model for prediction; step 2.5, judging the discontinuous points and extracting statistical characteristics after splicing the data in the cache region
Extracting break points according to the predicted different equipment states, wherein the break points are points of state change in the prediction result;
respectively processing according to whether the prediction result has a slot point:
1) if there is no break point, merging the prediction result with the data of the direct cache area and storing the data in the cache area;
2) if there is a break point, after merging the prediction result with the buffer area, sequentially traversing the prediction state of the segment formed by each break point, storing the obtained segment into a queue, and moving out the data before the next break point from the buffer area;
step 2.6 extraction of the processed signal segment
For the list of segments, the following statistical features of each segment are extracted: and (3) popping the sections from the queue in sequence according to the prediction states of the sections, the lengths and the mean values of the sections, judging whether the sections are a complete processing workpiece or not based on the statistical feature combination, wherein the judgment rule is a threshold value generated in the training stage, and the specific judgment rule is as follows:
1) the prediction state is a processing state, the length of the signal section is more than 10, the normalized mean value of the signal section is more than (max (the standard deviation of the mean value of the processing state of the training result-the training result is 1.5 times) and 0.5), the signal section is a normal processing state, the value of the workpiece counting variable is +1, and the information of the normal processing state is stored in a database;
2) if the prediction state is a processing state, the length of the signal section is less than 10, or the normalized mean value of the signal section is greater than (max (the training result processing state mean value-standard deviation of 1.5 times of the training result), 0.5), the prediction state is an abnormal processing state, and abnormal prompt needs to be performed on a user;
3) and in other cases, the next record is directly traversed in a non-processing state.
Step 2.7 parts replacement judgment and processing
If no new order data is transmitted, repeating the step 2.2 to the step 2.6; if a new order is introduced, all variables except work _ state in the predefined parameters in the step 2.1 are set to be null.
And continuously executing the flow until the device gateway is offline and no new data is transmitted in any more.
The above embodiments are illustrative of the present invention, and are not intended to limit the present invention, and any simple modifications of the present invention are within the scope of the present invention.
Claims (8)
1. A real-time statistical method for machine tool workpieces based on current signal segmentation is characterized in that: the method comprises the following steps:
a segmentation model training stage:
s11: collecting and processing data, namely collecting a historical current signal and historical work order information of a machine tool, and cleaning and normalizing the data to obtain a training sample;
s12: training a model, namely training a sample by using a GaussianHMM model, saving the trained model as a predictor, acquiring a segmentation point, and obtaining a segmented current signal; extracting statistical characteristics from the segmented current signals, setting a threshold value, and screening out workpiece processing signals;
the on-line segmentation and statistics stage of the workpiece:
s21: defining a cache region, and temporarily storing data in the cache region under the condition that no dividing point exists after the same order data is transmitted into a model for prediction; predefining other variables;
s22: acquiring current signals and work order number information acquired in real time in an interface mode;
s23: calling a pre-trained model, and transmitting cleaned data into model prediction; extracting break points according to different predicted states, wherein the break points are points of state change in a prediction result; acquiring a segment after current signal segmentation;
s24: sequentially traversing the segmentation segments, and extracting the statistical characteristics of each segment: the prediction state, the length and the mean value of each section are popped out from the queue in sequence, whether each section is a complete processing workpiece is judged based on the characteristic combination, and the judgment rule is a threshold value generated in the training stage;
s25: a change judgment and processing step of repeating steps S21 to S24 if no new order data is introduced; if a new order is introduced, some of the predefined variables of step S21 are set to null.
2. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 1, characterized in that: in the step S11, historical current signals and historical work order information of the machine tool are acquired through data acquired by the gateway, the frequency is 1S, the current signals comprise timestamps and current values, and the historical work order information comprises work order numbers, work order starting time and work order ending time.
3. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 2, characterized in that: before the data is cleaned in step S11, the two types of data are associated according to the time field, and the associated field is: time stamp, current value, work order number, work order start time, and work order end time.
4. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 3, characterized in that: the cleaning and normalization of the data in step S11 specifically includes the following steps:
s11.1 missing value processing, namely processing the time stamp into 1S intervals, and completing the missing current value in a backward filling mode;
s11.2 processing abnormal values, wherein the unified processing that the current value is less than 0 is 0; unifications with quantiles greater than 90% are filled with quantiles of 90%; the current value after treatment is stabilized in the range of 0-10A;
s11.3, normalization processing is carried out, wherein min-max normalization is adopted in the normalization processing mode;
s11.4, the data is deformed, and the shape of the input data is changed into a form of multiple rows and 1 column.
5. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 1, characterized in that: step S12 specifically includes the following steps:
s12.1 input data: inputting the current value of the data after completion of the cleaning in step S11;
s12.2 model selection: training input data by adopting a GaussianHMM model;
s12.3, setting model parameters: the states of the data analysis equipment at the early stage are 3 types: stopping, waiting and processing, so the number of the hidden states is set to be 3, the iteration turn is set to be 100, and other parameters are defaulted;
s12.4 model results are returned: after training, storing the model to the local; and returning the prediction result of the input data as the training result of the model;
s12.5, training result processing and processed workpiece statistics: splicing the input data and the training result, and finding out all break points according to the training result, wherein the break points are the first points of result state change; counting the average value and the length of each section according to the break points, wherein the equipment state corresponding to the average value which is the largest and the length which is greater than the average value is work _ state; the number of the processed workpieces can be counted according to the order number, the order time, the equipment state and the length.
6. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 1, characterized in that: before the step S23, the acquired signals are processed to construct the model-entering data, which includes filling of missing values, processing of transient abnormal values of the current, and normalization of the current data.
7. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 1, characterized in that: in step S23, according to whether there is a slot point in the prediction result, the following processing is performed:
a1. if there is no break point, merging the prediction result with the data of the direct cache area, and storing the data in the cache area;
a2. if there is a break point, the prediction result is merged with the buffer area, and then the prediction state of the segment formed by each break point is traversed in sequence, the obtained segment is stored in a queue, and the data before the next break point is shifted out from the buffer area.
8. A real-time statistical method for machine tool workpieces based on current signal segmentation as claimed in claim 1, characterized in that: the judgment rule in step S24 specifically includes the following steps:
b1. the prediction state is a processing state, the length of the signal section is more than 10, the normalized mean value of the signal section is more than (max (the standard deviation of the mean value of the processing state of the training result-the training result is 1.5 times) and 0.5), the signal section is a normal processing state, the value of the workpiece counting variable is +1, and the information of the normal processing state is stored in a database;
b2. if the prediction state is a processing state, the length of the signal section is less than 10, or the normalized mean value of the signal section is greater than (max (the training result processing state mean value-standard deviation of 1.5 times of the training result), 0.5), the prediction state is an abnormal processing state, and abnormal prompt needs to be performed on a user;
b3. otherwise, the next record is traversed directly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010855703.8A CN112171376B (en) | 2020-08-21 | 2020-08-21 | Machine tool workpiece real-time statistical method based on current signal segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010855703.8A CN112171376B (en) | 2020-08-21 | 2020-08-21 | Machine tool workpiece real-time statistical method based on current signal segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112171376A true CN112171376A (en) | 2021-01-05 |
| CN112171376B CN112171376B (en) | 2022-05-10 |
Family
ID=73924345
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010855703.8A Active CN112171376B (en) | 2020-08-21 | 2020-08-21 | Machine tool workpiece real-time statistical method based on current signal segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112171376B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113111854A (en) * | 2021-04-30 | 2021-07-13 | 平安国际融资租赁有限公司 | Current signal extraction method, current signal extraction device, computer equipment and storage medium |
| CN113361324A (en) * | 2021-04-25 | 2021-09-07 | 杭州玖欣物联科技有限公司 | Motor current anomaly detection method based on lstm |
| CN113554524A (en) * | 2021-07-29 | 2021-10-26 | 重庆允成互联网科技有限公司 | Intelligent production monitoring method and system based on industrial internet and storage medium |
| CN113762763A (en) * | 2021-09-01 | 2021-12-07 | 北京汽车集团越野车有限公司 | Engineering change data processing method and device and storage medium |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20010046482A (en) * | 1999-11-12 | 2001-06-15 | 이계철 | Method for choosing gaussian mixture number of hmm |
| US20050047646A1 (en) * | 2003-08-27 | 2005-03-03 | Nebojsa Jojic | System and method for fast on-line learning of transformed hidden Markov models |
| CN101059696A (en) * | 2007-04-20 | 2007-10-24 | 重庆大学 | Automatic collection method for mechanical processing task progress state based on machine tool power information |
| CN102254184A (en) * | 2011-07-18 | 2011-11-23 | 华中科技大学 | Method for fusing multi-physical-domain feature information |
| US20120303572A1 (en) * | 2011-05-24 | 2012-11-29 | Sony Corporation | Information processing apparatus, information processing method, and program |
| CN103105820A (en) * | 2012-05-22 | 2013-05-15 | 华中科技大学 | Machining cutter abrasion state identification method of numerical control machine tool |
| CN106181579A (en) * | 2016-08-23 | 2016-12-07 | 西安交通大学 | A kind of Tool Wear Monitoring method based on multisensor current signal |
| CN109396953A (en) * | 2018-12-05 | 2019-03-01 | 上海交通大学 | Lathe work condition intelligent identification system based on signal fused |
| CN110084301A (en) * | 2019-04-25 | 2019-08-02 | 山东科技大学 | A kind of multiple operating modes process industry and mining city method based on hidden Markov model |
| CN110347116A (en) * | 2019-07-17 | 2019-10-18 | 重庆大学 | A kind of conditions of machine tool monitoring system and monitoring method based on operation data stream |
| CN110400222A (en) * | 2019-07-26 | 2019-11-01 | 华南师范大学 | A kind of Stock Market Forecasting method based on hidden Markov and deep learning |
| CN110674752A (en) * | 2019-09-25 | 2020-01-10 | 广东省智能机器人研究院 | Hidden Markov model-based tool wear state identification and prediction method |
-
2020
- 2020-08-21 CN CN202010855703.8A patent/CN112171376B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20010046482A (en) * | 1999-11-12 | 2001-06-15 | 이계철 | Method for choosing gaussian mixture number of hmm |
| US20050047646A1 (en) * | 2003-08-27 | 2005-03-03 | Nebojsa Jojic | System and method for fast on-line learning of transformed hidden Markov models |
| CN101059696A (en) * | 2007-04-20 | 2007-10-24 | 重庆大学 | Automatic collection method for mechanical processing task progress state based on machine tool power information |
| US20120303572A1 (en) * | 2011-05-24 | 2012-11-29 | Sony Corporation | Information processing apparatus, information processing method, and program |
| CN102254184A (en) * | 2011-07-18 | 2011-11-23 | 华中科技大学 | Method for fusing multi-physical-domain feature information |
| CN103105820A (en) * | 2012-05-22 | 2013-05-15 | 华中科技大学 | Machining cutter abrasion state identification method of numerical control machine tool |
| CN106181579A (en) * | 2016-08-23 | 2016-12-07 | 西安交通大学 | A kind of Tool Wear Monitoring method based on multisensor current signal |
| CN109396953A (en) * | 2018-12-05 | 2019-03-01 | 上海交通大学 | Lathe work condition intelligent identification system based on signal fused |
| CN110084301A (en) * | 2019-04-25 | 2019-08-02 | 山东科技大学 | A kind of multiple operating modes process industry and mining city method based on hidden Markov model |
| CN110347116A (en) * | 2019-07-17 | 2019-10-18 | 重庆大学 | A kind of conditions of machine tool monitoring system and monitoring method based on operation data stream |
| CN110400222A (en) * | 2019-07-26 | 2019-11-01 | 华南师范大学 | A kind of Stock Market Forecasting method based on hidden Markov and deep learning |
| CN110674752A (en) * | 2019-09-25 | 2020-01-10 | 广东省智能机器人研究院 | Hidden Markov model-based tool wear state identification and prediction method |
Non-Patent Citations (3)
| Title |
|---|
| FENGQUAN ZHANG: "《A Gaussian mixture based hidden Markov model for motion recognition with 3D vision device》", 《COMPUTERS & ELECTRICAL ENGINEERING》 * |
| 周敏编著: "《制造业信息化工程学》", 31 January 2017 * |
| 方腾飞: "《基于负荷电流分析的多机床运行状态识别研究》", 《中国优秀硕士学位论文全文数据库工程科技I辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361324A (en) * | 2021-04-25 | 2021-09-07 | 杭州玖欣物联科技有限公司 | Motor current anomaly detection method based on lstm |
| CN113361324B (en) * | 2021-04-25 | 2023-06-30 | 杭州玖欣物联科技有限公司 | Lstm-based motor current anomaly detection method |
| CN113111854A (en) * | 2021-04-30 | 2021-07-13 | 平安国际融资租赁有限公司 | Current signal extraction method, current signal extraction device, computer equipment and storage medium |
| CN113554524A (en) * | 2021-07-29 | 2021-10-26 | 重庆允成互联网科技有限公司 | Intelligent production monitoring method and system based on industrial internet and storage medium |
| CN113554524B (en) * | 2021-07-29 | 2022-09-23 | 重庆允成互联网科技有限公司 | Intelligent production monitoring method and system based on industrial Internet and storage medium |
| CN113762763A (en) * | 2021-09-01 | 2021-12-07 | 北京汽车集团越野车有限公司 | Engineering change data processing method and device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112171376B (en) | 2022-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112171376B (en) | Machine tool workpiece real-time statistical method based on current signal segmentation | |
| CN104808587B (en) | A kind of mobility statistical method based on machining apparatus running status | |
| CN104794492A (en) | Online machine tool equipment machining and running state recognizing method based on power feature models | |
| CN109918817B (en) | Production line energy consumption analysis method based on time-varying data | |
| CN111832624A (en) | Tool remaining life prediction method based on anti-migration learning | |
| CN113253037B (en) | Current ripple-based edge cloud cooperative equipment state monitoring method and system and medium | |
| CN102609501B (en) | Data cleaning method based on real-time historical database | |
| CN116070527A (en) | Milling cutter residual life prediction method based on degradation model | |
| CN111993157A (en) | Machining equipment production state monitoring system and method based on electric signals | |
| CN106054832B (en) | Multivariable-based dynamic online monitoring method and device for intermittent chemical production process | |
| CN111176226A (en) | Automatic analysis method for alarm threshold of equipment characteristic parameter based on operation condition | |
| CN112416732B (en) | Hidden Markov model-based data acquisition operation anomaly detection method | |
| CN113722384A (en) | Detection method, system and equipment for abnormal value of measured point data based on density algorithm | |
| CN116859838B (en) | Early warning system for monitoring equipment operation condition | |
| CN111538755B (en) | Equipment operation state anomaly detection method based on normalized cross correlation and unit root detection | |
| CN115186935B (en) | Electromechanical device nonlinear fault prediction method and system | |
| CN117034197A (en) | Enterprise power consumption typical mode analysis method based on multidimensional Isolate-detection multi-point detection | |
| CN111599348A (en) | Automatic segmentation method and system for machine tool machining process monitoring signals | |
| CN112069168B (en) | Cloud storage method for equipment operation data | |
| CN111993158B (en) | Method for generating tool detection model, method, system, device and medium for detecting tool detection model | |
| CN114911786A (en) | Drilling depth curve optimization method, device, equipment and operation machine | |
| CN111076772B (en) | Processing method of cigarette shredding process data | |
| CN114326593B (en) | Cutter life prediction system and method | |
| CN116861204B (en) | Intelligent manufacturing equipment data management system based on digital twinning | |
| CN116787223A (en) | Cutter state prediction method of numerical control machine tool |
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 |