CN111428329B - Model-based machine learning system - Google Patents
Model-based machine learning system Download PDFInfo
- Publication number
- CN111428329B CN111428329B CN201811601457.2A CN201811601457A CN111428329B CN 111428329 B CN111428329 B CN 111428329B CN 201811601457 A CN201811601457 A CN 201811601457A CN 111428329 B CN111428329 B CN 111428329B
- Authority
- CN
- China
- Prior art keywords
- quality
- molding
- injection molding
- model
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000010801 machine learning Methods 0.000 title claims abstract description 43
- 238000000465 moulding Methods 0.000 claims abstract description 298
- 238000001746 injection moulding Methods 0.000 claims abstract description 154
- 238000012544 monitoring process Methods 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 47
- 238000004519 manufacturing process Methods 0.000 claims abstract description 42
- 238000004088 simulation Methods 0.000 claims abstract description 39
- 238000013500 data storage Methods 0.000 claims abstract description 12
- 230000007613 environmental effect Effects 0.000 claims abstract description 4
- 238000005457 optimization Methods 0.000 claims description 50
- 238000000034 method Methods 0.000 claims description 42
- 230000008569 process Effects 0.000 claims description 36
- 238000005259 measurement Methods 0.000 claims description 21
- 238000009826 distribution Methods 0.000 claims description 13
- 238000003908 quality control method Methods 0.000 claims description 9
- 238000013145 classification model Methods 0.000 claims description 8
- 230000002950 deficient Effects 0.000 claims description 7
- 238000007689 inspection Methods 0.000 claims description 5
- 230000006399 behavior Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000002787 reinforcement Effects 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 2
- 238000013139 quantization Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000007493 shaping process Methods 0.000 description 6
- 239000008186 active pharmaceutical agent Substances 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 238000002347 injection Methods 0.000 description 5
- 239000007924 injection Substances 0.000 description 5
- 238000011002 quantification Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000001816 cooling Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 239000002861 polymer material Substances 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Manufacturing & Machinery (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Injection Moulding Of Plastics Or The Like (AREA)
Abstract
The invention discloses a model-based machine learning system, which is used for calculating the optimal molding condition of injection molding and comprises a data storage device for providing a training data set; an injection molding program simulation unit for generating a group of simulation sensing data according to the input molding parameters; an injection molding process state monitoring unit for forming an injection molding process environment state according to the molding parameters, the group of analog sensing data and a molding quality state, wherein the molding quality state at least comprises a good product identification result; and the injection molding program optimizing unit comprises a molding parameter optimizer, trains the constructed molding parameter optimizing model according to the environmental state of the injection molding process, and leads the trained molding parameter optimizing model into the injection molding production line.
Description
Technical Field
The invention relates to a model-based machine learning system for calculating optimal molding conditions for injection molding.
Background
Injection molding is a complex process. Taking plastic injection molding as an example, the plastic injection molding is completed through a series of steps of plasticizing a polymer material, injecting the polymer material into a film cavity through pressure, compressing, cooling, ejecting and the like. There are many factors that affect the quality of injection molding. In practice, the plastic injection molding product is tested from the initial molding of the mold to stable mass production, and a series of molding parameter tests and adjustments are required to confirm that the parameters can stably produce the injection molding product meeting the design specifications of the product. Even if the molding parameters are calibrated, the molding quality is changed due to the production environment variation. In practice, the shaping parameters are adjusted and optimized by means of the experience of the personnel in the present stage to stabilize the shaping quality. However, the method of parameter adjustment is different, the training of experienced operators is not easy, the learning curve of operators on the novel injection molding equipment is not easy, and the defects of high personnel cost and difficult quality control are added, so that the method is one of important items to be solved in the molding and manufacturing industry.
In practice, the problems faced by the molding manufacturing industry include more complex product design, smaller molding window, improved molding environment influence on the quality of the product, and reduced molding stability and yield. Moreover, the customization degree of the current products is improved, the production frequency of the line change is improved due to a small number of various manufacturing modes, and a large amount of manpower is needed to optimize the molding parameters and stabilize the quality of the finished products, so that the manpower cost is greatly improved.
Taking a conventional injection molding process as an example, problems encountered in a molding parameter optimization method, such as difficulty in optimizing a plurality of acceptable conditions (the more complex the product design is, the smaller the molding window is, the more acceptable conditions are), a plurality of preset mark data capable of easily obtaining molding quantization quality are needed, however, the mark data are difficult to collect in practice. The difficulties encountered in conventional injection molding processes include: it is difficult to evaluate the molding parameters, and even an experienced molding process engineer cannot confirm the molding parameters empirically. Moreover, small and diverse manufacturing trends also make it difficult for the number of samples to accumulate in significant amounts to support traditional machine learning approaches. In addition, the quality index of the ejected product is often not easy to directly measure, such as burr degree, warping degree and the like. Even if the molding parameters are calibrated, the molding quality is changed due to the production environment variation. In the present stage, the shaping parameters are adjusted and optimized again by experienced personnel in practice, and the problems of high personnel cost, difficult quality control and management and the like are also caused.
Disclosure of Invention
The invention relates to a model-based machine learning system, which constructs a production environment variation model from historical data by introducing artificial intelligence technology, and automatically optimizes molding process parameters, so as to compensate quality variation caused by environment variation in real time.
According to one embodiment, a model-based machine learning system for calculating optimal molding conditions for injection molding is provided, comprising a data storage device for storing and processing data and providing a training data set; an injection molding program simulation unit for generating a group of simulation sensing data according to the input molding parameters; an injection molding process state monitoring unit for forming an injection molding process environment state according to the molding parameters, the group of analog sensing data and a molding quality state, wherein the molding quality state at least comprises a good product identification result; and an injection molding program optimizing unit, which adopts a molding parameter optimizer to train a built molding parameter optimizing model according to the environmental state of the injection molding process, and the trained molding parameter optimizing model is led into an injection molding production line.
For a better understanding of the above and other aspects of the invention, reference will now be made in detail to the following examples, which are illustrated in the accompanying drawings:
Drawings
FIG. 1 is a block diagram of a model-based machine learning system.
FIG. 2 is a block diagram of modeling and learning in a model-based machine learning system, according to an embodiment of the present invention.
FIG. 3 is a block diagram of an online learning by a model-based machine learning system, according to an embodiment of the present invention.
[ symbolic description ]
10 DS : data storage device
101: data preprocessing
D TD : training data set
100: actual injection molding manufacturing process
110: molding environment
200: injection molding program state monitoring unit
210: molded product detection system
220: external input unit
230: quantitative forming quality inference device
240: quality-changing shaping quality deducing device
250: quality acceptance discriminator
260: quantized molding quality data source selector
270: quality forming quality data source selector
280: module
MC: shaping parameters
D ES : analog sensing data
S k : environmental state of injection molding process
RE: rewards assessment unit
R: prize assessment
MC R : recommended molding parameters
MQ M : actual measurement data
MC A : actual forming parameters
D SD : actual sensed data
300: injection molding program optimizing unit
310: shaping parameter optimizer
400: injection molding program simulation unit
410: parameter statistics estimator
420: random data simulation generator
m: average value of
s: standard deviation of
Detailed Description
In the embodiment of the disclosure, a model-based machine learning system is provided for calculating the optimal molding condition of injection molding, so as to solve the difficulty in evaluating the molding parameters in the process of optimizing the molding parameters in practice, process big data required by the training stage of artificial intelligence technology, further consider the quality of molded products in real time and rapidly optimize in real time. Furthermore, the model-based machine learning system (including the simulation unit, the monitoring unit, the optimizing unit, and the estimators, the generators, the inference unit, the discriminant unit, the selector, the optimizer, etc. included in the foregoing units) according to the embodiments may be implemented by one or more logic operation units and/or processors. Examples of applications of the logic operation unit and/or the processor include, but are not limited to, a combination of one or more of a chip, a circuit board, and a recording medium storing a plurality of sets of program codes. Referring to fig. 1, a block diagram of a model-based machine learning system is shown. According to one embodiment of the present invention, an environment (environment) 10 (e.g., injection molding actual data) is provided by pre-designed experiments, such that an environment simulator (environment emulator) 14 obtains the appropriate amount of marking data from the environment (environment) 10 and constructs an operating environment model; by interaction with the environment model, the optimization program (optimization agent) 11 of the system can complete the learning process of optimization under the situation that the environment interaction cannot be obtained in real time. Furthermore, the optimizer initially completed in the optimizing process 11 can update timely according to the actual data accumulated in the actual injection molding process to interact with the environment 10 in real time for reinforcement learning. Therefore, by means of the model-based machine learning system, besides the adjustment of the molding process parameters of injection molding, the quality or characteristic value variation caused by the variation of molding environment can be compensated, so that the molding quality of the injection part can be optimized and stabilized, and the learning of the optimizer can be enhanced according to the accumulated actual data of the injection part according to the application requirement, so as to re-optimize the molding parameters.
Related embodiments are presented below, along with illustrations to detail the systems presented in this disclosure. The present disclosure is not limited to the units or devices illustrated in the system including the embodiments. Thus, the present disclosure does not show all possible examples, and other embodiments not presented in the present disclosure may also be applicable. Variations and modifications to the disclosed system may become apparent to those skilled in the art that do not necessarily depart from the spirit and scope of the disclosure. Accordingly, the specification and drawings are to be regarded in an illustrative manner and are not intended to limit the scope of the disclosure.
FIG. 2 is a block diagram of modeling and learning in a model-based machine learning system, according to an embodiment of the present invention. Furthermore, the system blocks shown in FIG. 2 may correspond to the flow of interaction between the environment simulator 14 and the optimization program 11 in FIG. 1.
As shown in FIG. 2, a model-based machine learning system for calculating optimal molding conditions for injection molding includes a data storage device 10 DS An injection molding process simulation unit 400 (Injection molding process emulator), an injection molding process state monitoring unit 200 (Injection molding process state observation unit), and an injection molding process optimization unit (Injection molding process optimization unit) 300.
In one embodiment, a Data storage device (Data storage) 10 DS For storing and processing data, wherein the stored raw production data (Production Raw Data) is subjected to a data preprocessing 101 to provide a training data set D TD . In one example, the production raw data includes, for example, a production molding number of actual injection molding, actual molding conditions, actual sensed data, and an actual product quality state. The product quality status includes, for example, a good product quality classification result (for example, classification of good product quality by True/False), and quality data of each acceptable condition. In one example, the quality data of the acceptable condition is, for example, injection molding flash, warpage, weight, size equalization data. Further, in an example, the data preprocessing 101 may include, but is not limited to, data screening, data merging, normalization, and the like.
In an embodiment, the injection molding process simulation unit 400 generates a set of simulated sensing data (emulated sensing data) D according to the molding parameters of the inputted molding conditions (molding condition) ES 。
In an embodiment, the injection molding process state monitoring unit 200 is based on the molding parameters MC and the analog sensing data D ES And a Quality State (QS) to form an injection molding process environment state (injection molding process state) S k Wherein the molding quality status at least includes a good product identification result.
Injection molding process optimizationThe unit 300 employs a molding parameter optimizer 310 (Injection molding condition optimizer) based on reinforcement learning algorithm, the molding parameter optimizer 310 based on the injection molding process environment state S k To train a constructed model of a model of parametric optimization (Molding condition optimization model). The trained molding parameter optimization model can be learned offline or imported into an injection molding production line for online learning.
The injection molding process simulation unit 400, the injection molding process state monitoring unit 200, and the injection molding process optimizing unit 300 are each further exemplified below.
In one example, the injection molding process state monitoring unit 200 includes at least one quality-of-molded quality estimator 240 (Acceptance state inference engine), the quality-of-molded quality estimator 240 based on the training data set D TD A good state classification model (Acceptance state classification model) is built. The quality-improving quality estimator 240 is based on the simulation sensing data D generated by the injection molding process simulation unit 400 according to the good product status classification model ES Deducing to deduce the set of analog sensing data D ES Quality of the molded product of (a). At this time, the molding quality status collected by the injection molding process status monitoring unit 200 includes good product identification results from the inference results of the quality inference unit 240. In an example, the quality modeling quality estimator 240 may be updated after each production run (but is not limited to).
Of course, in other examples, the injection molding process state monitoring unit 200 may further include a quantitative or other qualitative inference or selector, where the model built by the inference or selector may infer, for example, a quantitative quality result (such as the quantitative molding quality inference 230 proposed in the following examples), or a qualitative result that may determine whether to accept or not according to the inferred quantitative quality result (such as the quality acceptance identifier 250 or the qualitative molding quality data source selector 270 proposed in the following examples). Common intra-corollary modeling is, for example, classification by using a support vector classifier (Support Vector Classifier), linear discriminant (Linear Discriminant), nearest neighbor method (Nearest Neighbors), decision Tree (Decision Tree), random Forest (Random Forest), neural Network (Neural Network), and the like, but is not limited to the above.
In one example, the injection molding process state monitoring unit 200 further includes a quality molding quality data source selector 270 (Acceptance state input selector). The proof mass data source selector 270 builds a good identification inference model that determines the proof mass inference engine 240 for the set of simulated sensed data D ES After inference, the corresponding molded product belongs to good products or defective products in quality, and a molded item quality result is generated.
Furthermore, in an example, the injection Molding process state monitoring unit 200 further includes a quantification quality estimator 230 (Molding quality inference engine), so that the Molding quality state collected by the injection Molding process state monitoring unit 200 includes a quantification quality result (quantification quality) estimated by the quantification quality estimator 230 in addition to the good product identification result (at least from the result estimated by the quantification quality estimator 240). In an example, the quantitative modeling quality inference engine 230 may be updated after each production run (but is not limited to).
In one example, the quantitative modeling quality inference engine 230 builds a modeling project quantitative quality inference model (Molding quality inference model), for example, from a training data set. The quantitative modeling quality estimator 230 can estimate the model and the simulated sensing data D generated by the injection molding process simulation unit 400 according to the modeling item quantitative quality ES Deducing the set of analog sensing data D ES The quantitative quality results of the molded product.
In an example, the injection molding process state monitoring unit 200 may further include a quality enabling identifier 250 (Acceptance state identifier) and the quantized molding quality estimator 230, wherein the quality enabling identifier 250 performs quality determination on the quantized quality result from the quantized molding quality estimator 230. For example, if the burr value (quantized quality result) of the molded product deduced by the quantized molding quality deductor 230 is greater than 2mm, the quality acceptance criterion 250 determines that the quantized value of the item does not satisfy the acceptance criterion, and if the quantized value is less than or equal to 2mm, the quality acceptance criterion is determined to be satisfied. The quality acceptance arbiter 250 may set acceptance conditions for many different quantized items simultaneously. Therefore, the quality allowance discriminator 250 makes a quality determination based on the quantized result (the quantized result is used to infer the quality result). Furthermore, the quality allowance discriminator 250 may transmit the discriminated quality result to the quality molding quality data source selector 270, and the molded product corresponding to the quality discrimination result by the quality molding quality data source selector 270 is good or bad in quality. Thus, in this example, the molding quality status collected by the injection molding process status monitoring unit 200 may be obtained from the inference result of the quality inference unit 240 and the quality determination made by the quality acceptance determination unit 250 according to the quantization result (the inference result of the quantization molding quality inference unit 230), and selected by the quality data source selector 270.
Furthermore, in one example, the injection molding process state monitoring unit 200 may further include a module 280 coupled to the quality molding quality data source selector 270 and the molding parameter optimizer 310, respectively. After inference, the molding project quantified quality results, such as molding project measurement data MQ (Molding quality), from the molded product inferred by the quantified molding quality inference engine 230 are entered into the module 280 for compilation. The quality results derived from the quality inference unit 240 are processed by the quality data source selector 270 to generate quality results, such as good product identification results AS (acceptance state), for the molded product molding item, which are also entered into the module 280 for integration.
In an embodiment, the injection molding program simulation unit 400 can construct a correlation model of the molding parameters and the sensing data according to the history data of the actual process, and simulate the output of each sensing data of the molding according to the molding parameters inputted by the molding.
In one example, the injection molding program simulation unit 400 may be based on the phase of the training data set (actual data)Related parameters and data distribution, simulating simulated sensing data (emulated sensing data) D which is not real data according to the input molding parameters MC (molding condition) ES . Therefore, the injection molding process simulation unit 400 of the embodiment is configured such that the quality of the injection molding process simulation unit 400 is improved, or the combination of the quality of the injection molding process simulation unit 240 and the quality of the injection molding process simulation unit 230 is improved, not only by the quality of the training data set (actual data) or the quality and the quality of the training data set (actual data), but also by the simulation sensing data D generated by the injection molding process simulation unit 400 ES The (non-actual data) is subjected to the qualitative or qualitative and quantitative inference. Thus, the injection molding process simulation unit 400 of the embodiment may be configured to increase the data content (including data from actual or non-actual) obtained by the injection molding process state monitoring unit 200. One embodiment of the injection molding process simulation unit 400 is set forth below for illustration, but the disclosure is not limited thereto.
In one example, the injection molding process simulation unit 400 includes a parameter statistics estimator (Statistical parameter estimator) 410 and a random data simulation generator (Random number generator) 420. The parameter statistics estimator 410 may construct a correlation model based on actual molding parameters in the training data set and the respective actual sensed data distributions. The statistics of the respective simulated sensed data distributions corresponding to the simulated molding parameters are estimated from the simulated molding parameters MC of the molding conditions input to the injection molding program simulation unit 400, for example, based on the actual molding parameters in the training data set and the statistics of the respective actual sensed data distributions. The estimation method is, for example, interpolation (using, for example, nearest neighbor interpolation, linear interpolation, cubic interpolation or cubic interpolation (Cubic or Cubic Spline)), or other methods. In an example, the aforementioned statistics based on actual individual sensed data distributions may be the mean value m and standard deviation s of individual data statistics; the analog data is then estimated by suitable estimation calculations, such as interpolation or otherwise, and the statistics of each analog sensed data distribution, including the mean value m and standard deviation s of each data statistic, are estimated.
The random data simulation generator 420 may randomly generate a plurality of corresponding respective simulation sensing data according to the simulation molding parameters of the molding conditions input to the injection molding program simulation unit 400 based on the correlation model. Combining the corresponding analog sensing data to form a set of analog sensing data D ES To be provided to the injection molding process state monitoring unit (200). Wherein the random data simulation generator 420 randomly generates a plurality of corresponding respective simulated sensing data (e.g., simulated filling time) according to the statistics of the estimated respective sensing data distribution. Wherein the same analog molding parameter generates a plurality of different analog sensed data for the same sensed data item.
Therefore, the input and output simulated by the injection molding program simulation unit 400 are in one-to-many correspondence, i.e. the same molding parameters correspond to the same sensing data item, and different simulated sensing values are generated. In the injection molding process simulation unit 400 according to the embodiment, the correspondence between the analog input and the analog output is one-to-many, which is in accordance with the actual injection molding process. In an actual injection molding process, the same process parameters, but different sensing data (e.g., equipment sensing data and in-mold sensing features) may be generated.
According to the above illustration, the molding parameters MC, the analog sensed data D are input ES The molding item measurement data MQ (the quantized quality result of the molding item) derived from the quantized molding quality estimator 230, the good product identification result AS (the quantized quality result of the molding item, which may be derived from the quality estimator 240 and the quality admission discriminator 250 and passed through the quality data source selector 270), may be entered into the module 280 of the injection molding process state monitoring unit 200 for integration. Wherein the data deduced by the qualitative modeling quality inference engine 240 and the quality admission arbiter 250 may include simulated sensing data D from the training data set (history of actual processes) and the injection molding process simulation unit 400, respectively ES (historical data of non-actual processes).
Furthermore, in one example, the module 280 may also act as a driver (trigger) for the molding parameter optimizer. If the quality data source selector 270 determines that the molded product of the set of analog sensing data is good in quality, the module 280 outputs the injection molding process status with the molding order (e.g., outputs the injection molding process environment status S of the molding order) k To the molding parameter optimizer 310) to complete the last training of the molding parameter optimization model training process for the round, and to randomly pick a set of initial molding parameters again to perform the training process of the molding parameter optimization model for the next round, so that the molding parameter optimizer 310 continues to perform the training of the molding parameter optimization model. After a period of time, the injection molding process simulation unit 400 and the injection molding process state monitoring unit 200 may be updated according to the actual product quality results or by setting a predetermined time.
If the molding parameter optimization model has not completed the training process of the molding parameter optimization model for the round (i.e., the module 280 is driven to continue the optimization process), the molding parameter optimizer 310 may continue the training process of the molding parameter optimization model for the round according to the injection molding process state of the round. In one example, the molding parameter optimizer 310 may optimize the model according to the molding parameters of the mold number and the simulated injection molding process environment state S integrated by the injection molding process state monitoring unit 200 k Updating the molding parameter optimization model, recommending and inputting another set of molding conditions to the injection molding program simulation unit 400, and performing program simulation (through the injection molding program simulation unit 400) and program state monitoring (through the injection molding program state monitoring unit 200) of the next mold again until generating a set of optimized molding parameters to complete the training process of the molding parameter optimization model of the round. Details thereof are as described above and will not be repeated.
In this context, the training process of the molding parameter optimization model of each round is that after the molding parameter optimizer 310 initially sets a component type parameter in the injection molding process simulation unit 400, i.e. performs the parameter optimization process according to the existing parameter optimization model, if the module 280 determines that the component type parameter will generate a defective product, the molding parameter of another component type condition is recommended and input to the injection molding process simulation unit 400 until the recommended molding parameter enables the module 280 to determine that the simulated defective product can be generated, i.e. the parameter optimization model training of the round is completed. The molding parameter optimizer 310 then selects a new set of molding parameters for the next round of training of the parameter optimization model. Initially, it may take more, e.g., 20, recommendations and adjustments of molding parameters to complete the training process of the parametric optimization model for one round to allow the module 280 to determine that a simulated good product may be produced. However, as the number of rounds of training increases, the number of parameter adjustments required to complete each round decreases gradually (i.e., the number of parameter adjustments per round of training gradually converges), because it has been learned from past rounds of training how the corresponding molding parameter adjustments should be made for the injection molding conditions.
In addition, the user can selectively set the training of the molding parameter optimization model of the molding parameter optimizer 310 according to the practical application requirement, and can introduce the practical injection molding production line for use. For example, the total number of rounds R to be continuously performed may be set so that the number of parameter adjustment times for each round of completion training converges to a maximum of m rounds, and the number of rounds reaches n% or more of the total number of rounds, which may be regarded as preliminary completion of training of the molding parameter optimization model. R is, for example, the number of 10, 15, 20, 25, 30 rounds, or other rounds deemed appropriate by the user; m is, for example, 5, 4, 3, or other positive integer values as appropriate, n% is, for example, 80%, 85%, 90%, 95%, or other proportional values as appropriate, and the values of R, m, n are not particularly limited by the present disclosure. Taking r=20, m=5, n+=95% as an example, it is assumed that if training of the molding parameter optimization model is continuously performed for 20 rounds, the number of parameter adjustment times for each round of training is converged to at most 5 rounds to 95% or more of the total number of rounds, that is, the number of parameter adjustment times for 19 rounds is at most 5 (for example, all rounds including 5, 4, 3, 2, and 1 are calculated), the training of the molding parameter optimization model of the embodiment can be regarded as preliminary completion, and the actual injection molding production line can be introduced for use.
According to an embodiment, the molding parameter optimizer 310 of the injection molding process optimizing unit 300 constructs a molding parameter optimizing model including at least one molding process state and a plurality of sets of corresponding molding parameter adjusting actions, wherein the sets of corresponding relations are expected to yield expected values that allow molding quality to be achieved for the corresponding molding parameter adjusting actions under at least the inputted molding process state. A molding parameter optimization model of an embodiment that recommends optimized molding parameters is, for example, a neural network. Furthermore, the model of the molding parameter optimization may be updated automatically or by a user as needed, and the update frequency is not limited, and may be updated periodically or aperiodically, which is not limited in this disclosure.
In summary, in the embodiment, the training of the molding parameter optimizer 310 can be performed according to the good product state classification model and the molding item quality predictor model. Furthermore, in the embodiment, the bonus evaluation R (Reward Evaluation) (e.g. a bonus evaluation unit RE) may also be provided to the injection molding process optimization unit 300 according to the molding quality status collected by the injection molding process status monitoring unit 200. When the molding quality state is good (i.e., the good quality state is true), the reward is, for example, +1; when the molding quality state is defective (i.e., the good quality state is no), the prize is 0 or-1.
FIG. 3 is a block diagram of an online learning by a model-based machine learning system, according to an embodiment of the present invention. Furthermore, the system blocks shown in FIG. 3 may correspond to the flow of the optimizer 11 of FIG. 1 acting on the environment 10.
As shown in fig. 3, the injection molding line introduced by the model-based machine learning system of the embodiment includes an actual injection molding manufacturing process (Actual injection molding process) 100. Recommended molding parameters MC from the molding parameter optimizer 310 may be entered R (recommended molding conditions, a combination of molding parameters) in an actual injection molding process 100, wherein the actual injection molding process 100 isOutputting the actual molding parameters (applied molding condition) MC of the device A And actual sensed data D SD To the injection molding process state monitoring unit 200, and stored to the Data storage device (Data storage) 10 DS . The molding condition (combination of multiple molding parameters) has a time sequence, a sequence and a causal relation with the sensing data, namely, the molding parameters are causative (time sequence is prior), and the sensing data are the result (time sequence is later). The sensing data can be distinguished into equipment sensing related data, including molding equipment sensing data, peripheral equipment sensing data, mold sensing data and the like.
In one embodiment, the actual injection molding manufacturing process 100 is performed by a molding environment 110 including a series of activities such as molding parameter settings, injection molding, and producing molded products, wherein the molding environment 110 includes molding equipment, molds, and associated peripheral equipment or auxiliary systems such as mold warmers, dryers, cooling systems, and the like.
Furthermore, as described above with respect to the injection molding process state monitoring unit 200, it may include at least one quality modeling quality estimator 240. The quality estimator 240 shown in FIG. 3 is based on the established good state classification model (based on training data set or an updated data set) for the actual sensed data D outputted from the actual injection molding manufacturing process 100 SD To deduce that the actual sensing data D are obtained SD Quality of the molded product of (a). The injection molding process state monitoring unit 200 can integrate the molding quality state including at least a good product identification result for the actual injection molding manufacturing process. Alternatively, as described above, the injection molding process state monitoring unit 200 may further include a quantized estimator (such as the quantized molding quality estimator 230 set forth in the above example) or other quality estimators (such as the quality admission discriminator 250 set forth in the above example) that estimates the quality result based on the quantized result. The injection molding process state monitoring unit 200 may further include a selector (selector) associated with the quality/quantization; as shown in FIG. 3, the exemplary selectors include, for example, a quantized molding quality data source selector 270 and a quantized molding quality data source selector 260, each of which is rooted The result is selected and determined based on a plurality of different sources of the quality data for the texturized or quantized molding.
Furthermore, the model-based machine learning system of the embodiment may optionally further include a molded product detection system 210 (spot check update), which performs spot check on the actual product of the injection molding line, and performs actual measurement of the quality item (e.g., on the hardware equipment corresponding to the quality item) on the sampled product. Actual measurement of quality obtained by the molded product inspection system (e.g., molded item actual measurement data MQ M ) The quantized molding quality data source selector 260 may be transmitted to the injection molding process state monitoring unit 200. Thus, in this example, the quantized molding quality data source selector 260 aggregates the actual measurements of quality from the molded product detection system 210, and the quantized molding quality estimator 230 (which may be updated every time a mold) for the actual sensed data D SD Quantitative quality inference results of (2). Thus, in one example, the quantized molding quality data source selector 260 may determine the quantized quality of a molded product having these actual molding quality data based on a plurality of quantized molding quality data sources (e.g., 2 sources as shown in FIG. 3) and pass the molding item measurement data MQ (Molding quality) to the module 280. In one example, the priority of the source of the quantized molding quality data source selector 260 is, for example, the molded product detection system 210, or the quantized molding quality estimator 230. The present disclosure is not limited thereto.
As an example, the injection molding process state monitoring unit 200 may further include a quality allowance discriminator 250 for qualitatively discriminating the actual measurement result of quality collected from the quantized molding quality data source selector 260 from the quantized quality inference result. Accordingly, the quality allowance discriminator 250 may be configured to determine whether the molded product corresponding to the actual measurement results of the quality (the actual measurement results from the molded product detection system 210) and the molded product corresponding to the quantized quality inference results (from the quantized molding quality inference 230) collected by the quantized molding quality data source selector 260 are good or bad in quality.
Furthermore, the model-based machine learning system of the embodiment may optionally further include an external input unit 220 for directly inputting the acceptable quality result determined by the spot check of the actual product of the injection molding production line; for example, the inspector can directly observe and identify whether the molding products of the spot inspection are acceptable or not, and directly input the judging result into a processor. Thus, the external input unit 220 may also be referred to as an external quality molded quality data input unit (external acceptance state input unit). In one example, the acceptable quality results of the external input unit 220 are sent to the quality molding quality data source selector 270 of the injection molding process state monitoring unit 200. Thus, in one example, the source selector 270 may determine whether the molded product with the actual molding quality data is good or bad in quality based on the multiple sources (e.g., 3 sources as shown in fig. 3). As shown in fig. 3, the source of the qualitative forming quality data source selector 270 is, for example, the qualitative result deduced by the external input unit 220 (which may be updated every time a mode is executed), the qualitative result deduced by the qualitative forming quality deducing unit 240 (which may be updated every time a mode is executed), and the qualitative result judged by the quality allowance discriminator 250 (which deduces the qualitative result based on the quantitative result), which are selected and determined by the qualitative forming quality data source selector 270. The quality result (e.g., good product identification result AS) of the molded product molding item generated after passing through the quality molding quality data source selector 270 is entered into the module 280 for integration.
According to an embodiment, good product is identified as one of the quality molding qualities; in one example, the priority of the good product identification result AS is, for example, the external input unit 220, the quality control quality estimator 240, or the quality acceptance identifier 250. The present disclosure is not limited thereto.
Thus, as shown in FIG. 3, the module 280 of the injection molding process state monitoring unit 200 integrates the output device actual molding parameters MC A Actual sensed data D SD Molding project measurement data MQ (from quantized molding quality data source selector 260)Good product identification result AS (from the quality molding quality data source selector 270). Moreover, in the example, the module 280 may also be used AS a driver (trigger) of the molding parameter optimizer 310, that is, perform on-line learning (reinforcement learning, re-optimization) of the molding parameter optimizer 310 when the good product identification result AS is true (good product); otherwise, if the good product identification result AS is no (bad product), the molding parameters are optimized according to the injection molding program state of the mold number.
More specifically, in one example, if the quality data source selector 270 deduces that the molded product corresponding to the actual molding quality data is good in quality according to the quality approval results of the external input unit 220, the quality inference unit 240 and the quality approval discriminator 250, the module 280 stops triggering the molding parameter optimizer 310, and inputs the recommended molding conditions (molding parameters) to the next mold of the actual injection molding manufacturing process 100, and the molding parameter optimizer 310 performs incremental learning in batch on the injection molding line.
If the quality control data source selector 270 determines that the molded product corresponding to the actual molding quality data is defective in quality according to the acceptable quality results of the external input unit 220, the quality control estimator 240 and the quality acceptable discriminator 250, the module 280 triggers the molding parameter optimizer 310 to optimize the molding parameters. The molding parameter optimizer 310 can integrate the injection molding process environment state S according to the molding parameter optimization model and the injection molding process state monitoring unit 200 k Incremental learning of the molding parameter optimization model is performed (incremental learning). The molding parameter optimizer 310 recommends and inputs another set of molding conditions into the actual injection molding manufacturing process 100. Alternatively, depending on the actual application, the molding parameter optimizer 310 may perform training of the molding parameter optimization model again as shown in fig. 2 (training is as described above).
In addition, in an example, the good product identification result AS can be displayed in real time on the external input unit 220, and the user only needs to mark the wrong prediction result, so AS to reduce the operation load of the user. Furthermore, the quantized molding quality data can be displayed in real time on the external input unit 220, and the acceptable conditions input by the user are combined, so that the identification of the good product can be automatically determined to reduce the operation load of the user.
In summary, as shown in the embodiment system of fig. 3, the injection molding process state monitoring unit 200 mainly integrates the data generated by the actual injection molding manufacturing process 100 to describe the molding process environment state of the present mold. Thus forming process environment state S k Including the complete data generated by the actual injection molding manufacturing process 100, e.g., including the actual molding parameters MC A Actual sensed data D SD And a molded product quality status, wherein the molded product quality status may include at least a quality indicator, or both a quality indicator and a quantization indicator. In the embodiment, the quality index at least includes good quality identification result AS (based on quality data source of quality molding), and may further include quality results of other injection molding products, such AS binary classification results of whether there is a flow mark, whether there is a spray pattern, and the like. In an embodiment, the quantization index includes molding item measurement data MQ (based on a quantized molding quality data source), such as quantization data for flash length, finished weight, finished size, warpage level, or other product affecting factor items of the injection molded product. The molding parameter optimizer 310 determines the parameter adjustment behavior according to the inputted molding program state and generates a set of optimized molding parameters as the molding parameters of the next molding process. Furthermore, in one example, each of the history data of the mold-in-mold manufacturing is stored in the data storage device (production data storage portion) 10 DS And optionally to a central management system (Centralized Management System).
In addition, the embodiment provides that the molding parameter optimizer 310, the quality molding quality inference engine 240, the quality molding quality inference engine 230 and the like have corresponding inference models, and can be updated every time, and the update mechanism of these inference models is described as follows:
when the good product identification result is true (good product), the molding parameter optimizer 310 performs incremental learning of the molding parameter optimization model according to the batch parameter adjustment data;
the quantized molding quality inference engine 230 performs incremental learning of the quantized quality inference model based on the actual quantized quality measurement while the model provides the actual quantized quality measurement; and
the quality modeling quality estimator 240 performs incremental learning of the quality modeling model based on the actual quality measurements while the model number provides the actual quality measurements.
Furthermore, the quality data source selector 270 may perform incremental learning of the quality data source estimation model based on the actual quality data when the quality data source estimation model provides the actual quality data estimation result.
According to the above embodiment, the injection molding program simulation unit 400 can reduce the dependence of the parameter optimization learning process on the actual data, improve the use efficiency of the actual production data, and further improve the efficiency of the parameter optimization learning (simulate vs. actual injection). Furthermore, compared to the conventional molding parameter adjustment mode, the parameter adjustment mode of the molding parameter optimizer 310 of the embodiment can adjust a plurality of molding parameters for a plurality of optimization targets (quality inspection items, acceptable conditions) at the same time, so as to achieve the goal of optimizing the molding parameters, thus being a systematic and efficient parameter adjustment mode.
The injection molding process state monitoring unit 200 of the embodiment includes a molding quality estimator (e.g., the quality estimator 240, the quantized molding quality estimator 230) to construct the elements of the injection molding process state, which can reduce the need for tag data and assist in determining the timing of parameter optimization (e.g., the module 280 as a driver).
According to the above embodiments, a model-based machine learning system is provided, which utilizes the injection molding process simulation unit 400 to construct the state S of the injection molding process environment k A correlation model (molding parameter optimization model) with these molding parameter adjustment behaviors, and only a small amount of actual data is required for building The molding parameter optimization model is set, so that the actual data volume required by molding parameter optimization can be greatly reduced. Furthermore, as the number of training rounds of the molding parameter optimizer 310 increases, the number of parameter adjustments required for each round gradually decreases and converges to a minimum number. The system of the embodiment can thus quickly obtain the optimal molding conditions for injection molding. According to the test, the molding parameter optimizer 310 that completed learning displays: a probability of about 99.6% can accomplish parameter optimization in 3 modes (fig. 3). Preliminary verification, the model-based molding parameter optimizer as proposed in the embodiments can achieve a reduction in the number of test patterns required for the molding parameter optimization process. Therefore, the conventional injection process is manually adjusted by an experienced high-energy operator and mostly adjusts a single parameter at a time, and for the injection molding device of the system of the application embodiment, a plurality of allowable conditions can be optimized at the same time, so that the process time for searching for suitable molding parameters is greatly reduced, and a plurality of optimized molding parameters meeting the application conditions and requirements (such as different material properties of the manufactured products and different climatic conditions of the manufacturing place) can be obtained in a high-efficiency and real-time manner. When the embodiment is applied to a molding process with complex product design (smaller molding window and more acceptable conditions), the evaluation confirmation and efficiency of optimizing molding parameters are obviously improved. Therefore, the system of the embodiment has extremely high economic value and benefit in industrial application. In summary, the machine learning system based on the model provided by the embodiment can solve the difficulty in evaluating the molding parameter in the process of optimizing the molding parameter in practice, process big data required by the training stage of the artificial intelligence technology, and consider the molding quality in real time (the product quality can be known in real time, and real-time optimization can be performed rapidly).
The embodiments are presented in the drawings to describe one embodiment or application of the present disclosure, and the present disclosure is not limited to the scope and aspects of the illustrations. Other embodiments, such as combinations of different components or known components may be used, and adjustments and modifications may be made according to the needs of the actual application. The matter thus far described is illustrative only and not limiting. Those skilled in the art will recognize that, in the related system of the present disclosure, the molding parameters, measurement data, sensing data, admission items …, etc. to be optimized can be properly selected and adjusted according to the correlation factors affecting the practical application process, and the disclosure is not limited thereto.
In summary, although the present invention has been described in terms of the above embodiments, it is not limited thereto. Those skilled in the art will appreciate that various modifications and adaptations can be made without departing from the spirit and scope of the present invention. The scope of the invention is, therefore, indicated by the appended claims.
Claims (20)
1. A model-based machine learning system for calculating optimal molding conditions for injection molding, the model-based machine learning system comprising:
A data storage device for storing and processing data, wherein the data storage device provides a training data set after storing and processing raw data;
an injection molding program simulation unit for generating a group of simulation sensing data according to the input molding parameters;
the injection molding program state monitoring unit is used for forming an injection molding process environment state according to the molding parameters, the group of analog sensing data and the molding quality state, wherein the molding quality state at least comprises a good product identification result; and
an injection molding program optimizing unit which adopts a molding parameter optimizer based on a reinforcement learning algorithm, the molding parameter optimizer trains a constructed molding parameter optimizing model according to the environmental state of the injection molding process, the molding parameter optimizing model after training is led into an injection molding production line,
the injection molding program state monitoring unit comprises a quality modeling quality inference device, wherein the quality modeling quality inference device establishes a good state classification model according to the training data set, and the quality modeling quality inference device infers the group of analog sensing data generated by the injection molding program simulation unit according to the good state classification model so as to infer the quality of a molded product with the group of analog sensing data.
2. The model-based machine learning system of claim 1, wherein the injection molding program simulation unit comprises:
the parameter statistics estimator is used for constructing a relevance model according to actual forming parameters in the training data set and the distribution of each actual sensing data; and
a random data simulation generator randomly generates a plurality of corresponding respective simulation sensing data based on the correlation model according to a simulation molding parameter of the molding condition inputted to the injection molding program simulation unit.
3. The model-based machine learning system of claim 2, wherein the parameter statistics estimator of the injection molding process simulation unit estimates statistics of respective simulated sensed data distributions corresponding to the simulated molding parameters based on the actual molding parameters and the statistics of the actual respective sensed data distributions in the training dataset from simulated molding parameters of the molding conditions input to the injection molding process simulation unit, wherein the statistics of the actual respective sensed data distributions and the statistics of the respective simulated sensed data distributions each comprise an average (m) and a standard deviation(s) of the data statistics.
4. The model-based machine learning system of claim 3 wherein the random data simulation generator randomly generates a plurality of corresponding ones of the simulated sensed data based on statistics that estimate the distribution of the respective sensed data, wherein the same simulated molding parameters generate a plurality of different simulated sensed data for the same sensed data item.
5. The model-based machine learning system of claim 1, wherein the injection molding process state monitoring unit further comprises a quality data source selector that determines whether the molded product corresponding to the set of analog sensed data is good or bad in quality after the quality estimator estimates the set of analog sensed data.
6. The model-based machine learning system of claim 1, wherein the injection molding process state monitoring unit further comprises a quantitative molding quality estimator, the molding quality state assembled by the injection molding process state monitoring unit further comprising a quantitative quality result.
7. The model-based machine learning system of claim 6, wherein the quantitative modeling quality inference engine establishes a modeling project quantitative quality inference model based on the training data set, and the quantitative modeling quality inference engine infers the quantitative quality result of the molded product having the set of simulated sensing data based on a comparison of the modeling project quantitative quality inference model and the set of simulated sensing data generated by the injection molding process simulation unit.
8. The model-based machine learning system of claim 7, wherein the injection molding process state monitoring unit further comprises a quality acceptance determiner and the quantitative molding quality estimator, the quality acceptance determiner performing quality determination on the quantitative quality result from the quantitative molding quality estimator.
9. The model-based machine learning system of claim 8, wherein the injection molding process state monitoring unit further comprises a quality molding quality data source selector and the quality acceptance discriminator to determine whether the molded product corresponding to the quality discrimination result of the quality acceptance discriminator is good or bad in quality.
10. The model-based machine learning system of claim 5, wherein the injection molding process state monitoring unit further comprises a module coupled to the quality molding quality data source selector and the molding parameter optimizer, respectively,
if the modeling quality data source selector judges that the modeling product of the set of simulation sensing data belongs to good products in quality when the modeling parameter optimization model of one round is trained, the module outputs the injection molding process environment state to the modeling parameter optimizer, and the modeling parameter optimizer randomly selects a set of initial modeling parameters again to train the modeling parameter optimization model of the next round;
If the qualitative modeling quality data source selector determines that the molded product of the set of analog sensing data is defective in qualitative quality during the round, the module is driven to cause the molding parameter optimizer to continue training of the molding parameter optimization model for the round.
11. The model-based machine learning system of claim 1, wherein the molding parameter optimization model constructed by the molding parameter optimizer includes a plurality of sets of correspondences between at least one molding process state and corresponding molding parameter adjustment behaviors, the sets of correspondences being expected to yield expected values that allow molding quality for the corresponding molding parameter adjustment behaviors under at least the inputted molding process state.
12. The model-based machine learning system of claim 1, wherein the injection molding production line includes an actual injection molding manufacturing process, the recommended molding conditions from the molding parameter optimizer are input into the actual injection molding manufacturing process, and the actual injection molding manufacturing process outputs equipment actual molding parameters and actual sensing data to the injection molding process state monitoring unit.
13. The model-based machine learning system of claim 12, wherein the injection molding process state monitoring unit includes a quality inference engine that infers the quality of the molded product from the actual sensed data output by the actual injection molding manufacturing process based on the established good state classification model.
14. The model-based machine learning system of claim 13, wherein the injection molding process state monitoring unit further comprises a quality molding quality data source selector that determines whether the quality molding quality of the molded product with the actual sensing data is good or bad.
15. The model-based machine learning system of claim 14, further comprising an external input unit into which acceptable quality results obtained from spot check of actual products of the injection molding line are input, the acceptable quality results being transmitted to the quality data source selector.
16. The model-based machine learning system of claim 13, further comprising a molded product inspection system performing spot check and quality actual measurement on an actual product of the injection molding line, the injection molding process state monitoring unit further comprising a quantized molding quality estimator and a quantized molding quality data source selector, the actual injection molding manufacturing process further outputting actual sensed data to the quantized molding quality estimator, the quantized molding quality data source selector compiling comprising quality actual measurements of the molded product inspection system and quantized quality estimates of the actual sensed data by the quantized molding quality estimator.
17. The model-based machine learning system of claim 16, wherein the injection molding process state monitoring unit further comprises a quality acceptance determiner that qualitatively determines the quality of the actual quality measurements and the quantized quality inferences from the quantized molding quality data source selector.
18. The model-based machine learning system of claim 17, wherein the injection molding process state monitoring unit further comprises a quality molding quality data source selector that determines whether a molded product corresponding to a quality determination result of the quality acceptance determiner is good or bad in quality.
19. The model-based machine learning system of claim 18, further comprising an external input unit for inputting acceptable quality results from a spot check of an actual product of the injection molding line, the acceptable quality results being transmitted to the quality data source selector, wherein the injection molding process state monitoring unit further comprises a module coupled to the quality data source selector and the molding parameter optimizer, respectively,
If the quality data source selector judges that the quality of the molded product corresponding to the actual molding quality data is good according to the acceptable quality results of the external input unit and the quality judgment result of the quality acceptable judgment device, the module stops triggering the molding parameter optimizer and inputs the recommended molding condition to the next mold of the actual injection molding manufacturing process;
if the quality control quality data source selector deduces that the molded product corresponding to the actual molding quality data is determined to be defective in quality according to the quality control results of the external input unit and the quality control result of the quality control identifier, the module triggers the molding parameter optimizer to optimize molding parameters.
20. The model-based machine learning system of claim 1, wherein the trained molding parameter optimization model is imported into the injection molding line for on-line learning.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW107146534 | 2018-12-21 | ||
| TW107146534A TWI709922B (en) | 2018-12-21 | 2018-12-21 | A model-based machine learning system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111428329A CN111428329A (en) | 2020-07-17 |
| CN111428329B true CN111428329B (en) | 2024-01-05 |
Family
ID=71545610
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811601457.2A Active CN111428329B (en) | 2018-12-21 | 2018-12-26 | Model-based machine learning system |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111428329B (en) |
| TW (1) | TWI709922B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2022045698A (en) * | 2020-09-09 | 2022-03-22 | 株式会社日本製鋼所 | Machine learning method, computer program, machine learning device, and molding machine |
| TWI775207B (en) * | 2020-11-05 | 2022-08-21 | 財團法人資訊工業策進會 | Behavior recommendation apparatus, behavior recommendation method, and computer program product thereof |
| TWI787926B (en) * | 2021-07-30 | 2022-12-21 | 台達電子工業股份有限公司 | System and control method of compensating feedback control based on machine learning |
| WO2024118692A1 (en) * | 2022-11-30 | 2024-06-06 | iMFLUX Inc. | Universal process control for injection molding |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106393618A (en) * | 2015-07-28 | 2017-02-15 | 发那科株式会社 | Injection molding system for the calculation of optimum operating conditions and machine learning therefor |
| CN107263464A (en) * | 2016-04-04 | 2017-10-20 | 发那科株式会社 | Machine learning device, mechanical system, manufacture system and machine learning method |
| EP3248752A1 (en) * | 2016-05-27 | 2017-11-29 | Ashley Stone | Manufacturing process control systems and methods |
| CN108764597A (en) * | 2018-04-02 | 2018-11-06 | 华南理工大学 | A kind of product quality control method based on integrated study |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW567132B (en) * | 2000-06-08 | 2003-12-21 | Mirle Automation Corp | Intelligent control method for injection molding machine |
| TWI257894B (en) * | 2003-05-08 | 2006-07-11 | Ming-Shyan Huang | A cavity pressure based quality control method for injection molding process |
| US10733530B2 (en) * | 2016-12-08 | 2020-08-04 | Resurgo, Llc | Machine learning model evaluation in cyber defense |
| US10234848B2 (en) * | 2017-05-24 | 2019-03-19 | Relativity Space, Inc. | Real-time adaptive control of additive manufacturing processes using machine learning |
-
2018
- 2018-12-21 TW TW107146534A patent/TWI709922B/en active
- 2018-12-26 CN CN201811601457.2A patent/CN111428329B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106393618A (en) * | 2015-07-28 | 2017-02-15 | 发那科株式会社 | Injection molding system for the calculation of optimum operating conditions and machine learning therefor |
| CN107263464A (en) * | 2016-04-04 | 2017-10-20 | 发那科株式会社 | Machine learning device, mechanical system, manufacture system and machine learning method |
| EP3248752A1 (en) * | 2016-05-27 | 2017-11-29 | Ashley Stone | Manufacturing process control systems and methods |
| CN108764597A (en) * | 2018-04-02 | 2018-11-06 | 华南理工大学 | A kind of product quality control method based on integrated study |
Non-Patent Citations (3)
| Title |
|---|
| 杨金领 ; 尹自强 ; 关朝亮 ; 铁贵鹏 ; .基于支持向量机算法的微注射成型工艺参数优化.塑料工业.2009,(第05期),全文. * |
| 段尊雷 ; 任光 ; 张均东 ; 曹辉 ; .船舶机舱协作式模拟训练智能评估.交通运输工程学报.2016,(第06期),全文. * |
| 马俊燕 ; 廖小平 ; 夏薇 ; 莫玉军 ; .基于高斯过程机器学习的注塑过程建模及工艺参数优化设计.机械设计与制造.2013,(第03期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| TWI709922B (en) | 2020-11-11 |
| CN111428329A (en) | 2020-07-17 |
| TW202025005A (en) | 2020-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111428329B (en) | Model-based machine learning system | |
| US11403535B2 (en) | Model-based machine learning system | |
| JP6294268B2 (en) | Abnormality diagnosis device for injection molding machine | |
| Sun et al. | A hybrid approach to cutting tool remaining useful life prediction based on the Wiener process | |
| KR102139358B1 (en) | Process control method using machine learning-based platform, computer program medium for performing same, and process control appratus | |
| CN111461555A (en) | Production line quality monitoring method, device and system | |
| CN110571792A (en) | Analysis and evaluation method and system for operation state of power grid regulation and control system | |
| CN117274259B (en) | Keyboard production equipment management method and system based on Internet of things | |
| CN117008557B (en) | Production control method and system for blending type interpenetrating network thermoplastic elastomer | |
| KR20210091130A (en) | Methods and systems for improving physical production processes | |
| CN113435699A (en) | Intelligent quality control method and system | |
| KR102366922B1 (en) | heavy electric equipment deterioration prediction system based on time-series isolation diagnostic information and method therefor | |
| CN111491773B (en) | Molding parameter processing method for injection press | |
| CN114692507A (en) | Counting data soft measurement modeling method based on stacking Poisson self-encoder network | |
| KR102470836B1 (en) | Artificial intelligence based continuous process control device, method of predicting quality and improving yield using the same | |
| US20200086546A1 (en) | Method for controlling film production | |
| CN117669384A (en) | Intelligent monitoring method and system for temperature sensor production based on Internet of things | |
| CN110851920B (en) | Automatic generation method of main reinforcement line of die material pressing device | |
| CN112801558B (en) | Optimization method and device of process parameter adjustment action decision model | |
| Friederich et al. | A Framework for Validating Data-Driven Discrete-Event Simulation Models of Cyber-Physical Production Systems | |
| CN116414095A (en) | Data-driven optimization method for technological parameters in traditional Chinese medicine manufacturing process | |
| CN110610016A (en) | Method for predicting rail transit stopping problem based on big data machine learning | |
| CN111144639A (en) | Subway equipment fault prediction method and system based on ALLN algorithm | |
| Xin et al. | Dynamic probabilistic model checking for sensor validation in Industry 4.0 applications | |
| CN113049793A (en) | Production detection verification method and device for Abs composite material |
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 |