CN110059738A - A kind of method for early warning and system of quality of die casting - Google Patents
A kind of method for early warning and system of quality of die casting Download PDFInfo
- Publication number
- CN110059738A CN110059738A CN201910289983.8A CN201910289983A CN110059738A CN 110059738 A CN110059738 A CN 110059738A CN 201910289983 A CN201910289983 A CN 201910289983A CN 110059738 A CN110059738 A CN 110059738A
- Authority
- CN
- China
- Prior art keywords
- data
- quality
- injection
- speed
- curve
- 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
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B22—CASTING; POWDER METALLURGY
- B22D—CASTING OF METALS; CASTING OF OTHER SUBSTANCES BY THE SAME PROCESSES OR DEVICES
- B22D17/00—Pressure die casting or injection die casting, i.e. casting in which the metal is forced into a mould under high pressure
- B22D17/20—Accessories: Details
- B22D17/32—Controlling equipment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- 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
A kind of method for early warning of quality of die casting, comprising the following steps: the injection curve data of acquisition die-casting process process in real time, including ram position curve, injection speed curve and injection pressure curve；Extract at a slow speed injection speed, quick injection speed, turn at a slow speed quick position, pressurization settling time, maximum speed, maximum pressure as product quality characteristic data carry out 0-1 normalized generation Product Process feature vector, with current mould time number binding, as product quality data characteristics vector input quality discrimination module, there is the support vector machine classifier obtained based on historical data training；Quality discrimination module is handled Product Process feature vector by support vector machine classifier, to judge whether Product Process data characteristics is located in the multidirectional quantity space region of product qualification, triggering alarm when the judgment is no.The present invention can identify real-time, accurately and efficiently the abnormal quality of die casting, and greatly reduce subsequent testing cost.
Description
Technical field
The present invention relates to die casting proceeding in quality control fields, a kind of method for early warning more particularly to quality of die casting and are
System.
Background technique
Compression casting technology is in a kind of production efficiency and all higher casting method of mechanization degree and modern manufacturing
Develop a kind of faster forming technique cut less.Compression casting refers to liquid metal under high speed, high pressure, is injected into accurate add
In the die cast metal type of work and under stress crystallization and freezing, clear-cut, any surface finish the die casting of formation.In recent years, with
The degree of automation of the continuous development of technology, die casting machine is also higher and higher, and the information content that entire die casting production line is covered also is got over
Come more.But the information management and processing means for die casting production line are but far from satisfying the need of modern Die Casting Enterprise
It asks.
Statistical Process Control (spc) is a kind of process control tool by mathematical statistics method.It to production process into
Row assay finds the sign that system sexual factor occurs in time according to feedback information, and takes measures to eliminate its influence, makes
Journey maintains the slave mode only influenced by random factor, to achieve the purpose that control quality.It has been applied to much press now
Cast enterprise, however be limited to die casting machine automatization level and die casting environment it is severe, traditional control chart method is difficult in reality
It is used in the production of border.
For Modern Manufacturing Enterprise, continuous improvement with client to quality requirement, in addition computer, database and
The fast development of industrial network technology causes product quality tracing system to become one indispensable in quality control system
Point, efficient creation data automatic collection, statistics, production procedure control may be implemented and learn by functions such as product inquiries, but
It can not automatically extract out and obtain relationship and rule in data, lack the method for mining data behind knowledge.
Summary of the invention
It is a primary object of the present invention to overcome the deficiencies of the prior art and provide a kind of method for early warning of quality of die casting and
System.
To achieve the above object, the invention adopts the following technical scheme:
A kind of method for early warning of quality of die casting, comprising the following steps:
The injection curve data of S1, in real time acquisition die-casting process process, the injection curve data include ram position song
Line, injection speed curve and injection pressure curve；
S2, according to the injection curve data, by will at a slow speed injection stage and quick injection stage take mean value respectively, obtain
To injection speed at a slow speed and quick injection speed, and on ram position curve, determination turns at a slow speed quick position, in injection speed
It writes music and respectively obtains maximum speed and maximum pressure on line and injection pressure curve, then will injection speed, quick injection speed at a slow speed
Degree turns at a slow speed quick position, pressurization settling time, maximum speed, maximum pressure as product quality characteristic data progress 0-1
Normalized generates Product Process feature vector, and with current mould time number binding, together as product quality data characteristics to
Input quality discrimination module is measured, wherein the quality discrimination module has the historical data training based on the injection curve data
Obtained support vector machines (SVM) classifier；
S3, the quality discrimination module by the support vector machines (SVM) classifier to Product Process feature vector into
Row processing, to judge whether process data feature possessed by product is located in the multidirectional quantity space region of product qualification, when sentencing
Break triggering alarm when being no.
Further:
Historical data training based on the injection curve data obtains the process packet of support vector machines (SVM) classifier
It includes:
(1) it is split processing using time series data of the sliding window algorithm to die casting curve, in position curve
Ram position is start position as judgment basis, and time series data is divided into [lts/lw] section, wherein ltsIt is long for total time series data
Degree, lwFor productive temp time, the time series data being so just divided into total data in single productive temp time；
(2) data after segmentation are subjected to feature extraction, for taking mean value to obtain slowly respectively with the quick injection stage at a slow speed
Fast injection speed and quick injection speed, and found in corresponding ram position curve and turn quick position at a slow speed, and in injection speed
It writes music line and injection pressure curve respectively obtains maximum speed and maximum pressure；Each characteristic value is subjected to 0-1 normalization, conversion
Function is as follows:
Wherein max is the maximum value of sample data, and min is the minimum value of sample data, and x is handled specific features
Value, x*For the characteristic value after normalization；Thus dimensionless feature feature vector being normalized between 0-1；
(3) quality of die casting information database is recalled, in order binds quality of die casting information and Product Process feature,
And data are upset into rearrangement, it is divided into test data set and training dataset, and utilize and construct based on support vector machines (SVM)
Sorting algorithm chooses Radial basis kernel function RBF as kernel function, and RBF function representation is as follows:
Wherein k represents RBF function, X1With X2Two different characteristic value vectors are respectively represented, σ indicates sample variance；
(4) training dataset is trained, obtains SVM classifier；
(5) SVM classifier that training obtains is tested in test data set, if success rate prediction is met the requirements,
Then SVM classifier building is completed, if not satisfied, then return course step (4), overline adjusting parameter are trained.
Training dataset is trained using the method for following stochastic gradient descent in step (4), wherein every time repeatedly
Generation all random selection learning samples update model parameter, disposably handle all sample bring memory overheads to reduce, more
New rule is as follows:
Wherein:
w(i+1)For the targeted parameter value after i-th iteration；
w(i)Indicate the parameter value that the (i-1)-th iteration obtains；
XiIndicate sample；
γ is learning rate, represents iteration step length,The operator for seeking gradient is represented, J is cost function, specific as follows:
Wherein m is sample size,For the predicted value for each sample, yiFor the occurrence of each sample, if the sample
Die casting is then 1 there are quality problems, is then 0 there is no quality problems, and by iteration, J function drops to whole when tending towards stability
Only train.
A kind of early warning system of quality of die casting, it is pre- in die-casting process process progress quality using the method for early warning
It is alert.
Further, the early warning system of the quality of die casting includes data capture unit, data processing unit and matter
Measure prewarning unit；The data capture unit obtains injection curve data from for executing the die casting unit that die-casting process operates,
Injection curve data, production batch, mould time and the technological parameter information of setting are stored in product database, and data are believed
Breath is sent to the data processing unit；The data processing unit is by the data obtained dividing processing, according to the injection curve
Data, by will at a slow speed injection stage and quick injection stage take mean value respectively, obtain injection speed at a slow speed and quick injection speed
Degree, and determination turns at a slow speed quick position on ram position curve, on injection speed curve and injection pressure curve respectively
Obtain maximum speed and maximum pressure, then will at a slow speed injection speed, quick injection speed, turn quick position at a slow speed, pressurization is built
Between immediately, maximum speed, maximum pressure as product quality characteristic data carry out 0-1 normalized generate Product Process feature
Vector, and bound with current mould time number, product quality data characteristics vector is obtained, the quality pre-alert unit is passed to；It is described
Quality pre-alert unit is handled Product Process feature vector by support vector machines (SVM) classifier, to judge product institute
Whether the process data feature having is located in the multidirectional quantity space region of product qualification, when the judgment is no triggering alarm.
The data obtained feature is input in SVM classifier by the quality pre-alert unit, judges whether product quality is qualified,
If product is unqualified, alarm, and indicate specific product batch, go wrong and die casting machine number, pass through work later
Industry internet is fed back to and the operating side of the die casting unit, and qualified just default enters next mould, and gained is judged and produced
Product characteristic is according in current production mould time number deposit database.
Further include the data storage cell, is used for data capture unit, data processing unit and quality pre-alert list
The resulting data deposit database of member is saved as historical data.
It further include server and remote human-machine's interactive unit, the server includes monitoring module and internet interface,
Described monitor conveys data by the internet interface after module passes through the data of partition identification distinct device and stored；
Received data integration is sent to remote human-machine's interactive unit by interchanger by the data capture unit；It is described long-range
Man-machine interaction unit includes client and dynamic monitoring interface module, and the client includes memory module and communication module.
The invention has the following beneficial effects:
The method for early warning and system of quality of die casting proposed by the present invention, the production number of more die casting units of real-time reception
According to, therefrom extract at a slow speed injection speed, quick injection speed, turn at a slow speed quick position, pressurization settling time, maximum speed,
Maximum pressure as product quality characteristic data carry out 0-1 normalized generate Product Process feature vector, and with current mould
Secondary number binding, obtains product quality data characteristics vector, using the product quality data characteristics vector extracted, is gone through using basis
The history creation data support vector machines classifier that training is established in advance carrys out the quality judging of real-time perfoming product, can be timely
The product quality problem for reliably finding die casting, avoids heavy losses, needs testing cost after reduction, instruct subsequent production.This
The die casting product processing quality method of discrimination proposed is invented, it is more superior relative to conventional statistics quality control chart method, it can
Abnormal quality that is more acurrate, more efficiently identifying die casting, greatly reduces subsequent testing cost, significantly improves economy
Benefit.In addition, early warning system can by the way of the creation data of long-range real-time reception die casting unit, and early warning system with
Remote monitoring system module is connected, and is no longer limited by the limitation of workshop locating for die casting unit, can remotely identify Product Process
There is abnormal die casting unit in information, and then convenient for management, improves production efficiency.
Detailed description of the invention
Fig. 1 is the establishment process of the quality of die casting differentiation SVM classifier in an embodiment of the present invention.
Fig. 2 is the early warning system block diagram of the quality of die casting of an embodiment of the present invention.
Fig. 3 is the long-range monitor client human-computer interaction interface based on an embodiment of the present invention.
Specific embodiment
It elaborates below to embodiments of the present invention.It is emphasized that following the description is only exemplary,
The range and its application being not intended to be limiting of the invention.
Refering to fig. 1, in one embodiment, a kind of method for early warning of quality of die casting, comprising the following steps:
The injection curve data of S1, in real time acquisition die-casting process process, the injection curve data include ram position song
Line, injection speed curve and injection pressure curve；
S2, according to the injection curve data, by will at a slow speed injection stage and quick injection stage take mean value respectively, obtain
To injection speed at a slow speed and quick injection speed, and on ram position curve, determination turns at a slow speed quick position, in injection speed
It writes music and respectively obtains maximum speed and maximum pressure on line and injection pressure curve, then will injection speed, quick injection speed at a slow speed
Degree turns at a slow speed quick position, pressurization settling time, maximum speed, maximum pressure as product quality characteristic data progress 0-1
Normalized generates Product Process feature vector, and with current mould time number binding, together as product quality data characteristics to
Input quality discrimination module is measured, wherein the quality discrimination module has the historical data training based on the injection curve data
Obtained support vector machines (SVM) classifier；
S3, the quality discrimination module by the support vector machines (SVM) classifier to Product Process feature vector into
Row processing, to judge whether process data feature possessed by product is located in the multidirectional quantity space region of product qualification, when sentencing
Break triggering alarm when being no.
In method for early warning of the invention, by the creation data of more die casting units of real-time reception, therefrom extracts and press at a slow speed
Firing rate degree, quick injection speed turn at a slow speed quick position, pressurization settling time, maximum speed, maximum pressure as product matter
Measure feature data carry out 0-1 normalized and generate Product Process feature vector, and bind with current mould time number, obtain product matter
Amount data characteristics vector is trained using the product quality data characteristics vector extracted using according to historical production data in advance
The support vector machines classifier of foundation carrys out the quality judging of real-time perfoming product, can timely and reliably find die casting
Product quality problem avoids heavy losses, needs testing cost after reduction, instruct subsequent production.Die casting proposed by the present invention produces
Product processing quality method of discrimination, it is more superior relative to conventional statistics quality control chart method, it can more acurrate, more efficiently identify
The abnormal quality of die casting out greatly reduces subsequent testing cost, significantly improves economic benefit.
In a preferred embodiment, the historical data training based on the injection curve data obtains support vector machines
(SVM) process of classifier includes:
(1) it is split processing using time series data of the sliding window algorithm to die casting curve, in position curve
Ram position is start position as judgment basis, and time series data is divided into [lts/lw] section, wherein ltsIt is long for total time series data
Degree, lwFor productive temp time, the time series data being so just divided into total data in single productive temp time；
(2) data after segmentation are subjected to feature extraction, for taking mean value to obtain slowly respectively with the quick injection stage at a slow speed
Fast injection speed and quick injection speed, and found in corresponding ram position curve and turn quick position at a slow speed, and in injection speed
It writes music line and injection pressure curve respectively obtains maximum speed and maximum pressure；Each characteristic value is subjected to 0-1 normalization, conversion
Function is as follows:
Wherein max is the maximum value of sample data, and min is the minimum value of sample data, and x is handled specific features
Value, x*For the characteristic value after normalization；Thus dimensionless feature feature vector being normalized between 0-1；
(3) quality of die casting information database is recalled, in order binds quality of die casting information and Product Process feature,
And data are upset into rearrangement, it is divided into test data set and training dataset, and utilize and construct based on support vector machines (SVM)
Sorting algorithm chooses Radial basis kernel function RBF as kernel function, and RBF function representation is as follows:
Wherein k represents RBF function, X1With X2Two different characteristic value vectors are respectively represented, σ indicates sample variance；
(4) training dataset is trained, obtains SVM classifier；
(5) SVM classifier that training obtains is tested in test data set, if success rate prediction is met the requirements,
Then SVM classifier building is completed, if not satisfied, then return course step (4), overline adjusting parameter are trained.
In a more preferred embodiment, the method in step (4) using following stochastic gradient descent is to training dataset
It is trained, wherein each iteration, which all randomly chooses learning sample, updates model parameter, disposably handles all samples to reduce
It is as follows to update rule for bring memory overhead:
Wherein:
w(i+1)For the targeted parameter value after i-th iteration；
w(i)Indicate the parameter value that the (i-1)-th iteration obtains；
XiIndicate sample；
γ is learning rate, represents iteration step length,The operator for seeking gradient is represented, J is cost function, specific as follows:
Wherein m is sample size,For the predicted value for each sample, yiFor the occurrence of each sample, if the sample
Die casting is then 1 there are quality problems, is then 0 there is no quality problems, and by iteration, J function drops to whole when tending towards stability
Only train.
In another embodiment, a kind of early warning system of quality of die casting is added using the method for early warning in die casting
Work process carries out quality pre-alert.
In a preferred embodiment, the early warning system of the quality of die casting includes data capture unit, data processing list
Member and quality pre-alert unit；The data capture unit obtains injection song from for executing the die casting unit that die-casting process operates
Injection curve data, production batch, mould time and the technological parameter information of setting are stored in product database by line number evidence, and
The data processing unit is sent by data information；The data processing unit is by the data obtained dividing processing, according to described
Injection curve data, by will at a slow speed injection stage and quick injection stage take mean value respectively, obtain injection speed at a slow speed and fast
Fast injection speed, and determination turns at a slow speed quick position on ram position curve, it is bent in injection speed curve and injection pressure
Respectively obtain maximum speed and maximum pressure on line, then will at a slow speed injection speed, quick injection speed, turn quick position at a slow speed
It sets, be pressurized settling time, maximum speed, maximum pressure as the progress 0-1 normalized generation production of product quality characteristic data
Product technology characteristics vector, and bound with current mould time number, product quality data characteristics vector is obtained, the quality pre-alert is passed to
Unit；The quality pre-alert unit is handled Product Process feature vector by support vector machines (SVM) classifier, to sentence
Whether process data feature possessed by stopping pregnancy product is located in the multidirectional quantity space region of product qualification, triggers when the judgment is no
Alarm.
In a preferred embodiment, the data obtained feature is input in SVM classifier by the quality pre-alert unit, judgement
Whether product quality is qualified, if product is unqualified, alarms, and indicates specific product batch, goes wrong and die casting
Machine number is fed back to by industry internet later and the operating side of the die casting unit, qualified just default enters next mould,
And gained judgement is produced in mould time number deposit database with product feature data according to current.
In a preferred embodiment, the early warning system of quality of die casting further includes the data storage cell, for that will count
It is protected according to acquiring unit, data processing unit and the resulting data deposit database of quality pre-alert unit as historical data
It deposits.
In a preferred embodiment, the early warning system of quality of die casting further includes server and remote human-machine's interactive unit,
The server includes monitoring module and internet interface, and the module of monitoring passes through the data of partition identification distinct device simultaneously
Data are conveyed by the internet interface after being stored；The data capture unit passes through interchanger for received data set
At being sent to remote human-machine's interactive unit；Remote human-machine's interactive unit includes OPC client and dynamic monitoring interface
Module, the OPC client include memory module and communication module.
The feature and advantage of the specific embodiment of the invention are further described below in conjunction with attached drawing.
Fig. 2 be quality of die casting early warning system logic diagram, the quality of die casting of a specific embodiment it is real-time pre-
Alert system includes: data capture unit, data processing unit, data storage cell, quality pre-alert unit.
Data capture unit obtains die-casting process data from the die casting unit that die-casting process operates is executed.
The die casting unit includes die casting machine, soup feeding machine device people, spray robot, pickup robot.The die casting unit
Connect industrial personal computer.The opc server and and industrial computer that the industrial personal computer includes industrial computer, connect with industrial computer
The die casting cell controller of connection；The controller is connect by fieldbus with all PLC, for pressing described in collection in worksite
The die casting information of unit is cast, for industrial computer for receiving, handling the die casting information in the controller, die-cast product is basic
The data such as information such as injection curve are acquired by industrial computer and are shown.Data communication can be realized by ModBus bus.Pass through institute
State controller the die casting unit is controlled, Guan Suoshu industrial personal computer.The opc server includes monitoring module and interconnection
Network interface, the monitoring module are defeated by the internet interface after passing through the data of partition identification distinct device and being stored
Send mass data；Received data integration is sent to remote human-machine's interactive unit by interchanger；Remote human-machine's interaction
Unit includes OPC client and dynamic monitoring interface module, and the OPC client includes memory module and communication module.
The data capture unit includes: comprehensive wiring system and OPC client, and affiliated comprehensive wiring system includes handing over
It changes planes, Shielded Twisted Pair, unshielded twisted pair, the Shielded Twisted Pair carries out network for every die casting unit and affiliated workshop
Transmission, an end is connected to the opc server network interface, another interchanger being terminated between per car；The unshielded twisted pair
It is connect for interchanger with the OPC client of centrally located control room；The OPC client to every workshop carry out remote monitoring with
The process data that each die casting unit provides in real time is obtained, and condition discrimination and prewarning unit are obtained according to the affiliated judgment model of data
Required data；
The data processing unit, connect with OPC client, a large amount of untreated technique information data is obtained, by institute
The data of acquisition are split arrangement, and combine demand and corresponding algorithm, generate maximum die casting speed, maximum die casting pressure etc.
Product Process feature is shown and is stored in monitoring interface；
The quality pre-alert unit, is integrated in dynamic monitoring module, communicates with data processing unit, by data processing unit
Obtained product quality characteristic carries out product quality judgement by SVM (support vector machines) algorithm, and is shown in monitoring interface
Show, alarms the situation of product processing quality exception.The data storage cell is communicated with remote monitoring system, by institute
It states data capture unit and determines to be saved with prewarning unit the data obtained deposit database as historical data with state.
During the work time, remote monitoring system opc server monitors die casting unit industrial computer, OPC service first
Production technology information is transmitted to OPC client by industry internet by device, and at OPC client, data capture unit is by technique
Production information includes that fast curve, pressure curve and production batch, mould time and the technological parameter information of setting is pressed to be stored in product
In database, while data processing unit is sent by product data.Data processing unit can pass through data preprocessing procedures
It realizes, by ram position curve, fast curve, pressure curve is pressed to do according to the information such as product batches number and pitch time, by gained
Data dividing processing, at a slow speed injection speed, quick injection speed, turn quick position, pressurization settling time, maximum speed at a slow speed
Degree, maximum pressure, and quality pre-alert unit is passed data to, the data obtained feature is input to classification and calculated by quality pre-alert unit
In method, judge whether process data feature possessed by product is located in the multidirectional quantity space region of product qualification to get production is arrived
Quality whether just alarm at control interface, and indicate specific product batch, gone out if product is unqualified by He Ge judgement
Existing problem and die casting machine number are fed back to by industry internet later and die casting unit operating side, qualified just default enter
Next mould, and gained judgement is produced in mould time number deposit database with product feature data according to current.
Fig. 3 is the main modular of long distance monitoring client end interface, including production information interface, real time monitoring interface and alarm
Interface.The production information interface, for showing that zero class of every die casting machine current output, the current output of day shift, the middle class in a kindergarten are worked as
Preceding output, day Current production, cumulative production；The real time monitoring interface, for showing the production phase locating for every die casting machine,
Including the sealing of hole stage presses at a slow speed in feeding stage, die casting stage, spraying stage, pickup stage, trimming stage and injection process
Penetrate commencing speed, maximum speed, average speed and the end speed of formed punch；Commencing speed, the maximum of packing stage injection punch head
Speed, average speed and end speed；And increase the stroke and end position of the injection punch head in stage；The alert interface,
Production real time status including more die casting machines, if production status is well just shown in green, there is exception and just becomes in the quality of production
Alarm date, time of fire alarming, variable name and variable description when showing that injection exception occurs in die casting machine for red and pop-up.
The above content is combine it is specific/further detailed description of the invention for preferred embodiment, cannot recognize
Fixed specific implementation of the invention is only limited to these instructions.For those of ordinary skill in the art to which the present invention belongs,
Without departing from the inventive concept of the premise, some replacements or modifications can also be made to the embodiment that these have been described,
And these substitutions or variant all shall be regarded as belonging to protection scope of the present invention.
Claims (8)
1. a kind of method for early warning of quality of die casting, which comprises the following steps:
The injection curve data of S1, in real time acquisition die-casting process process, the injection curve data include ram position curve, pressure
Penetrate rate curve and injection pressure curve；
S2, according to the injection curve data, by will at a slow speed injection stage and quick injection stage take mean value respectively, obtain slow
Fast injection speed and quick injection speed, and determination turns at a slow speed quick position on ram position curve, in injection speed song
Respectively obtain maximum speed and maximum pressure on line and injection pressure curve, then will injection speed, quick injection speed, slow at a slow speed
Speed turns quick position, pressurization settling time, maximum speed, maximum pressure as product quality characteristic data and carries out 0-1 normalizing
Change processing and generate Product Process feature vector, and is bound with current mould time number, it is defeated as product quality data characteristics vector together
Enter quality discrimination module, wherein there is the quality discrimination module historical data training based on the injection curve data to obtain
Support vector machines (SVM) classifier；
S3, the quality discrimination module by the support vector machines (SVM) classifier to Product Process feature vector at
Reason, to judge whether process data feature possessed by product is located in the multidirectional quantity space region of product qualification, when being judged as
Alarm is triggered when no.
2. the method for early warning of quality of die casting as described in claim 1, which is characterized in that based on the injection curve data
Historical data training obtains the process of support vector machines (SVM) classifier and includes:
(1) it is split processing using time series data of the sliding window algorithm to die casting curve, with pressure head in position curve
Position is start position as judgment basis, and time series data is divided into [lts/lw] section, wherein ltsFor total time series data length,
lwFor productive temp time, the time series data being so just divided into total data in single productive temp time；
(2) data after segmentation are subjected to feature extraction, for taking mean value to be pressed at a slow speed respectively with the quick injection stage at a slow speed
Firing rate degree and quick injection speed, and found in corresponding ram position curve and turn quick position at a slow speed, and in injection speed song
Line and injection pressure curve respectively obtain maximum speed and maximum pressure；Each characteristic value is subjected to 0-1 normalization, transfer function
It is as follows:
Wherein max is the maximum value of sample data, and min is the minimum value of sample data, and x is handled specific features value, x*For
Characteristic value after normalization；Thus dimensionless feature feature vector being normalized between 0-1；
(3) quality of die casting information database is recalled, in order binds quality of die casting information and Product Process feature, and will
Data upset rearrangement, are divided into test data set and training dataset, and utilize and construct classification based on support vector machines (SVM)
Algorithm chooses Radial basis kernel function RBF as kernel function, and RBF function representation is as follows:
Wherein k represents RBF function, X1With X2Two different characteristic value vectors are respectively represented, σ indicates sample variance；
(4) training dataset is trained, obtains SVM classifier；
(5) SVM classifier that training obtains is tested in test data set, if success rate prediction is met the requirements, SVM
Classifier building is completed, if not satisfied, then return course step (4), overline adjusting parameter are trained.
3. the method for early warning of quality of die casting as claimed in claim 2, which is characterized in that in step (4) using it is following with
The method of machine gradient decline is trained training dataset, wherein each iteration all randomly chooses learning sample more new model ginseng
Number disposably handles all sample bring memory overheads to reduce, it is as follows to update rule:
Wherein:
w(i+1)For the targeted parameter value after i-th iteration；
w(i)Indicate the parameter value that the (i-1)-th iteration obtains；
XiIndicate sample；
γ is learning rate, represents iteration step length, and ▽ represents the operator for seeking gradient, and J is cost function, specific as follows:
Wherein m is sample size,For the predicted value for each sample, yiFor the occurrence of each sample, if the sample die casting
Part is then 1 there are quality problems, is then 0 there is no quality problems, and by iteration, J function, which drops to, terminates instruction when tending towards stability
Practice.
4. a kind of early warning system of quality of die casting, which is characterized in that use early warning as described in any one of claims 1 to 3
Method carries out quality pre-alert in die-casting process process.
5. the early warning system of quality of die casting as claimed in claim 4, which is characterized in that including data capture unit, data
Processing unit and quality pre-alert unit；The data capture unit is obtained from for executing the die casting unit that die-casting process operates
Injection curve data, production batch, mould time and the technological parameter information of setting are stored in product database by injection curve data
In, and the data processing unit is sent by data information；The data processing unit by the data obtained dividing processing, according to
The injection curve data, by will at a slow speed injection stage and quick injection stage take mean value respectively, obtain injection speed at a slow speed
With quick injection speed, and on ram position curve determination turn quick position at a slow speed, in injection speed curve and injection pressure
Respectively obtain maximum speed and maximum pressure on force curve, then will at a slow speed injection speed, quick injection speed, turn quick at a slow speed
Position, pressurization settling time, maximum speed, maximum pressure carry out the generation of 0-1 normalized as product quality characteristic data
Product Process feature vector, and bound with current mould time number, product quality data characteristics vector is obtained, it is pre- to pass to the quality
Alert unit；The quality pre-alert unit is handled Product Process feature vector by support vector machines (SVM) classifier, with
Judge whether process data feature possessed by product is located in the multidirectional quantity space region of product qualification, touches when the judgment is no
Hair alarm.
6. the early warning system of quality of die casting as claimed in claim 5, which is characterized in that the quality pre-alert unit is by gained
Data characteristics is input in SVM classifier, whether qualified judges product quality, if product is unqualified, is alarmed, and indicates have
Body product batches go wrong and die casting machine number, are fed back to later by industry internet and the die casting unit
Operating side, qualified just default enters next mould, and gained judgement is deposited with product feature data according to current production mould time number
Enter in database.
7. such as the early warning system of the described in any item quality of die casting of claim 4 to 6, which is characterized in that further include the number
According to storage unit, for data capture unit, data processing unit and the resulting data of quality pre-alert unit to be stored in data
Library is saved as historical data.
8. such as the early warning system of the described in any item quality of die casting of claim 4 to 7, which is characterized in that further include server
With remote human-machine's interactive unit, the server includes that monitoring module and internet interface, the monitoring module pass through subregion
Data are conveyed by the internet interface after identifying the data of distinct device and being stored；The data capture unit passes through
Received data integration is sent to remote human-machine's interactive unit by interchanger；Remote human-machine's interactive unit includes client
End and dynamic monitoring interface module, the client includes memory module and communication module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910289983.8A CN110059738A (en) | 2019-04-11 | 2019-04-11 | A kind of method for early warning and system of quality of die casting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910289983.8A CN110059738A (en) | 2019-04-11 | 2019-04-11 | A kind of method for early warning and system of quality of die casting |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110059738A true CN110059738A (en) | 2019-07-26 |
Family
ID=67317587
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910289983.8A Pending CN110059738A (en) | 2019-04-11 | 2019-04-11 | A kind of method for early warning and system of quality of die casting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059738A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111177655A (en) * | 2019-12-30 | 2020-05-19 | 联想(北京)有限公司 | Data processing method and device and electronic equipment |
| CN111368416A (en) * | 2020-03-02 | 2020-07-03 | 三鑫重工机械有限公司 | Large-scale steel casting mold filling sensing correction method and system |
| CN114612262A (en) * | 2022-04-28 | 2022-06-10 | 成都秦川物联网科技股份有限公司 | Intelligent manufacturing industrial Internet of things system and control method |
| CN114632922A (en) * | 2022-02-15 | 2022-06-17 | 湖南大学 | High-efficiency high-quality aluminum alloy damping tower high-pressure vacuum die-casting process control system and method |
| CN115041544A (en) * | 2022-08-11 | 2022-09-13 | 深圳市信润富联数字科技有限公司 | Method and device for detecting abnormality of stamping part and storage medium |
| CN117371626A (en) * | 2023-12-07 | 2024-01-09 | 小米汽车科技有限公司 | Casting quality prediction method, device and medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101786150A (en) * | 2010-02-10 | 2010-07-28 | 广东伊之密精密机械有限公司 | Die casting machine and automatic judgment method of die casting quality |
| CN102218522A (en) * | 2011-06-03 | 2011-10-19 | 柯有权 | Quality comprehensive control system for die casting machine |
| CN103411974A (en) * | 2013-07-10 | 2013-11-27 | 杭州赤霄科技有限公司 | Cloud big data-based planar material detection remote system and cloud big data-based planar material detection method |
| JP2014069219A (en) * | 2012-09-28 | 2014-04-21 | Direct 21 Corp | Monitor display means, molding apparatus using the same and monitor display method |
| CN106600140A (en) * | 2016-12-07 | 2017-04-26 | 河海大学常州校区 | Improved support vector machine-based gas pipeline fault prediction and early warning system and method |
| CN107093175A (en) * | 2017-04-21 | 2017-08-25 | 华南师范大学 | A kind of electrokinetic cell welding quality detection method |
-
2019
- 2019-04-11 CN CN201910289983.8A patent/CN110059738A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101786150A (en) * | 2010-02-10 | 2010-07-28 | 广东伊之密精密机械有限公司 | Die casting machine and automatic judgment method of die casting quality |
| CN102218522A (en) * | 2011-06-03 | 2011-10-19 | 柯有权 | Quality comprehensive control system for die casting machine |
| JP2014069219A (en) * | 2012-09-28 | 2014-04-21 | Direct 21 Corp | Monitor display means, molding apparatus using the same and monitor display method |
| CN103411974A (en) * | 2013-07-10 | 2013-11-27 | 杭州赤霄科技有限公司 | Cloud big data-based planar material detection remote system and cloud big data-based planar material detection method |
| CN106600140A (en) * | 2016-12-07 | 2017-04-26 | 河海大学常州校区 | Improved support vector machine-based gas pipeline fault prediction and early warning system and method |
| CN107093175A (en) * | 2017-04-21 | 2017-08-25 | 华南师范大学 | A kind of electrokinetic cell welding quality detection method |
Non-Patent Citations (8)
| Title |
|---|
| FANGYIN CHENG AND YAN MA: "Quality Control System for Die Casting Based on Statistic Analysis", 《APPLIED MECHANICS AND MATERIALS》 * |
| 东北工学院宫克强: "《特种铸造》", 31 January 1982, 机械工业出版社 * |
| 冯兴东: "《分布式统计计算》", 30 April 2018, 上海财经大学出版社 * |
| 孙志敏 等: "《铝以及铝合金加工技术》", 31 December 2013, 冶金工业出版社 * |
| 张弛 等: "压铸单元远程监控系统的开发与应用", 《特种铸造及有色合金》 * |
| 李健: "《匝道交通控制理论与方法》", 31 July 2013, 北京交通大学出版社 * |
| 钱庆 等: "基于支持向量机的拉弯工件弯曲质量检测研究", 《制造技术与机床》 * |
| 黄浩凯 等: "基于安卓的压铸车间监控系统的设计", 《特种铸造及有色合金》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111177655A (en) * | 2019-12-30 | 2020-05-19 | 联想(北京)有限公司 | Data processing method and device and electronic equipment |
| CN111177655B (en) * | 2019-12-30 | 2022-04-22 | 联想(北京)有限公司 | Data processing method and device and electronic equipment |
| CN111368416A (en) * | 2020-03-02 | 2020-07-03 | 三鑫重工机械有限公司 | Large-scale steel casting mold filling sensing correction method and system |
| CN114632922A (en) * | 2022-02-15 | 2022-06-17 | 湖南大学 | High-efficiency high-quality aluminum alloy damping tower high-pressure vacuum die-casting process control system and method |
| CN114612262A (en) * | 2022-04-28 | 2022-06-10 | 成都秦川物联网科技股份有限公司 | Intelligent manufacturing industrial Internet of things system and control method |
| CN114612262B (en) * | 2022-04-28 | 2022-07-22 | 成都秦川物联网科技股份有限公司 | Intelligent manufacturing industry Internet of things system and control method |
| US11619928B1 (en) | 2022-04-28 | 2023-04-04 | Chengdu Qinchuan Iot Technology Co., Ltd. | Industrial internet of things system for smart manufacturing and control method thereof |
| US11953886B2 (en) | 2022-04-28 | 2024-04-09 | Chengdu Qinchuan Iot Technology Co., Ltd. | Industrial internet of things system for intelligent production control and control method thereof |
| CN115041544A (en) * | 2022-08-11 | 2022-09-13 | 深圳市信润富联数字科技有限公司 | Method and device for detecting abnormality of stamping part and storage medium |
| CN117371626A (en) * | 2023-12-07 | 2024-01-09 | 小米汽车科技有限公司 | Casting quality prediction method, device and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059738A (en) | A kind of method for early warning and system of quality of die casting | |
| US11860610B2 (en) | Systems of industrial internet of things for industrial production control and control methods thereof | |
| CN101950327A (en) | Equipment state prediction method based on fault tree information | |
| CN106250288A (en) | Root alarm analysis and identification method based on data mining | |
| CN1987697A (en) | Fault diagnostic system and method for under industrial producing process small sample condition | |
| CN114118673A (en) | Workshop intelligent fault diagnosis early warning method based on digital twin technology | |
| CN106845826B (en) | PCA-Cpk-based cold continuous rolling production line service quality state evaluation method | |
| CN110689642B (en) | Abnormal driving distinguishing method and system based on vehicle-mounted OBD data and probability statistics | |
| US20210065021A1 (en) | Working condition state modeling and model correcting method | |
| CN111390648B (en) | Turning tool abrasion judging method based on antagonistic neural network | |
| CN114897102A (en) | Industrial robot fault diagnosis method, system, equipment and storage medium | |
| CN106054832B (en) | Multivariable-based dynamic online monitoring method and device for intermittent chemical production process | |
| CN106199421A (en) | A kind of method for early warning based on the big data of industry and system | |
| CN108268018A (en) | A kind of machine-building Workshop Production progress information acquisition system based on RFID | |
| CN113130014B (en) | Rare earth extraction simulation method and system based on multi-branch neural network | |
| CN116703254B (en) | Production information management system for mechanical parts of die | |
| CN109034540A (en) | A kind of lathe serial arrangement dynamic prediction method based in article process stream | |
| CN111906772B (en) | Intelligent product processing method based on industrial robot | |
| CN105913142A (en) | Method for improving accuracy of order completion time through using workshop RFID data | |
| CN110598750B (en) | Working condition identification method based on switch machine action curve similarity characteristics | |
| CN100440089C (en) | Industrial process nonlinear fault diagnosis system and method based on fisher | |
| CN108052087A (en) | Manufacturing process multivariate quality diagnostic classification device based on comentropy | |
| CN105260948B (en) | A kind of water-supply systems daily planning scheduling decision method | |
| CN106777672A (en) | Metal material forging microstructure flexible measurement method based on adaptive expert system | |
| CN112836577B (en) | Intelligent traffic unmanned vehicle fault gene diagnosis method and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190726 |
|
| RJ01 | Rejection of invention patent application after publication |