CN113762366B - Method and system for predictive control of forming state of additive manufacturing - Google Patents
Method and system for predictive control of forming state of additive manufacturing Download PDFInfo
- Publication number
- CN113762366B CN113762366B CN202110973673.5A CN202110973673A CN113762366B CN 113762366 B CN113762366 B CN 113762366B CN 202110973673 A CN202110973673 A CN 202110973673A CN 113762366 B CN113762366 B CN 113762366B
- Authority
- CN
- China
- Prior art keywords
- forming state
- prediction
- forecast
- additive manufacturing
- state label
- 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
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 75
- 239000000654 additive Substances 0.000 title claims abstract description 73
- 230000000996 additive effect Effects 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012545 processing Methods 0.000 claims abstract description 12
- 230000004927 fusion Effects 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 21
- 230000015654 memory Effects 0.000 claims description 18
- 238000013527 convolutional neural network Methods 0.000 claims description 14
- 239000010410 layer Substances 0.000 claims description 11
- 230000007547 defect Effects 0.000 claims description 10
- 239000002356 single layer Substances 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 7
- 239000011324 bead Substances 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 239000002893 slag Substances 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000006403 short-term memory Effects 0.000 abstract description 3
- 230000007787 long-term memory Effects 0.000 abstract description 2
- 238000003466 welding Methods 0.000 description 15
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000011217 control strategy Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 238000010146 3D printing Methods 0.000 description 1
- 208000019454 Feeding and Eating disease Diseases 0.000 description 1
- 208000018460 Feeding disease Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
- B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
- B29C64/00—Additive manufacturing, i.e. manufacturing of three-dimensional [3D] objects by additive deposition, additive agglomeration or additive layering, e.g. by 3D printing, stereolithography or selective laser sintering
- B29C64/30—Auxiliary operations or equipment
- B29C64/386—Data acquisition or data processing for additive manufacturing
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
- B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
- B29C64/00—Additive manufacturing, i.e. manufacturing of three-dimensional [3D] objects by additive deposition, additive agglomeration or additive layering, e.g. by 3D printing, stereolithography or selective laser sintering
- B29C64/30—Auxiliary operations or equipment
- B29C64/386—Data acquisition or data processing for additive manufacturing
- B29C64/393—Data acquisition or data processing for additive manufacturing for controlling or regulating additive manufacturing processes
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B33—ADDITIVE MANUFACTURING TECHNOLOGY
- B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- B33Y50/00—Data acquisition or data processing for additive manufacturing
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B33—ADDITIVE MANUFACTURING TECHNOLOGY
- B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- B33Y50/00—Data acquisition or data processing for additive manufacturing
- B33Y50/02—Data acquisition or data processing for additive manufacturing for controlling or regulating additive manufacturing processes
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Materials Engineering (AREA)
- Chemical & Material Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Manufacturing & Machinery (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mechanical Engineering (AREA)
- Optics & Photonics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for predictive control of an additive manufacturing forming state, which belong to the field of additive manufacturing control, and comprise the following steps: collecting video data of an additive manufacturing forming area and inputting the video data into a forecast model frame by frame; the method comprises the steps of extracting features of each frame of image input into a prediction model by using an encoder, outputting the extracted features, and predicting the output of the encoder in sequence by using a long-term and short-term memory network module to obtain prediction feature information in the next period; classifying the prediction characteristic information by using a classifier to obtain a first prediction forming state label, reconstructing the prediction characteristic information by using a decoder to obtain a molten pool image in the next period, and performing image processing on the molten pool image to obtain a second prediction forming state label; and fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label so as to control the additive manufacturing strategy of the next period. Various quality problems in additive manufacturing are accurately predicted, and accurate control is realized.
Description
Technical Field
The invention belongs to the field of additive manufacturing control, and in particular relates to a method and a system for predicting and controlling an additive manufacturing forming state.
Background
Additive manufacturing is a technique of systematically stacking raw materials to obtain solid parts, mainly slicing and stacking based on an existing three-dimensional model, and completing a printing process layer by layer, thus also being called 3D printing. In recent years, additive manufacturing has attracted attention from more and more researchers due to its great advantages of low cost and high efficiency, and various additive manufacturing techniques have been developed and gradually expanded from nonmetallic parts to metallic part forming. However, most of the current additive manufacturing technology development is still under exploration, and the parts obtained by direct forming are easy to have internal defects or morphology defects such as cracks, pores, unfused, abnormal accumulation and the like, so that the yield is low, and the high-efficiency low-cost manufacturing cannot be realized. Accordingly, there is a need for rapid diagnostics and corresponding control of the forming quality in additive manufacturing of formed parts, thereby ensuring defect free manufacture of the parts.
The existing additive manufactured part quality control method generally processes images, and does not fully utilize continuous video data in the forming process, so that deviation of later forming prediction is caused. In addition, the output is mostly quality grade when the existing method diagnoses the forming quality, only can roughly judge whether the forming quality has a problem, can not accurately describe the type of the quality problem, is inconvenient for subsequent process adjustment, and is unfavorable for realizing high-quality additive manufacturing.
Disclosure of Invention
Aiming at the defects and improvement demands of the prior art, the invention provides a method and a system for predicting and controlling the forming state of additive manufacturing, which aim to accurately predict various quality problems in the additive manufacturing and realize accurate control.
To achieve the above object, according to one aspect of the present invention, there is provided an additive manufacturing forming state prediction control method including: s1, collecting video data of an additive manufacturing forming area and inputting the video data into a prediction model frame by frame, wherein the prediction model comprises an encoder, a long-period and short-period memory network module, a classifier and a decoder; s2, extracting features of each frame of image input into the prediction model by using the encoder, outputting the extracted features, and carrying out sequence prediction on the output of the encoder by using the long-period and short-period memory network module to obtain prediction feature information in the next period; s3, classifying the prediction feature information by using the classifier to obtain a first prediction forming state label, reconstructing the prediction feature information by using the decoder to obtain a molten pool image in the next period, and performing image processing on the molten pool image to obtain a second prediction forming state label; s4, fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label, and controlling an additive manufacturing strategy of the next period according to the final forecast forming state label.
Furthermore, the first forecast forming state label and the second forecast forming state label both comprise a plurality of sub-labels and occurrence probability of each sub-label, and the plurality of sub-labels comprise normal, path error, unstable welding machine process, welding wire shake, wire feeding unsmooth, welding bead flowing, excessive slag and shutdown.
Still further, fusing the first forecast forming status tag and the second forecast forming status tag in S4 includes: and respectively fusing the same sub-labels in the first forecast forming state label and the second forecast forming state label, and setting the sub-label with the highest occurrence probability after fusion as the final forecast forming state label.
Further, the fusion in S4 is an average fusion, a maximum fusion, a weight fusion or a generalized average fusion.
Still further, the step S1 further includes: carrying out additive manufacturing experiments by adopting different technological parameters for a plurality of times to obtain various defects in each forming state, and collecting video data of an additive manufacturing forming area in the experimental process; sampling the video data to obtain an image set, dividing the image set into a plurality of groups of subsets by taking a preset length N and a preset step length as sliding windows, and labeling x images in each subset to obtain a labeling data set, wherein x is less than N/3; the predictive model is trained with the top N-x images in each of the subsets as input and the annotation dataset as output.
Still further, the training the predictive model includes: and training the forecasting model by using a Adam optimization algorithm and taking the minimum loss function of the forecasting model as a training target, wherein the loss function adopts an average absolute error.
Furthermore, the encoder in S2 performs feature extraction on each frame of image input into the prediction model through a single-layer convolutional neural network, a multi-layer convolutional neural network, a single-layer convolutional neural network or a multi-layer convolutional neural network, and outputs the extracted frame of image.
Further, the decoder in S3 reconstructs the prediction characteristic information through a reverse network corresponding to the encoder.
According to another aspect of the present invention, there is provided an additive manufacturing forming state predictive control system comprising: the system comprises an acquisition module, a prediction model and a prediction module, wherein the acquisition module is used for acquiring video data of an additive manufacturing forming area and inputting the video data into the prediction model frame by frame, and the prediction model comprises an encoder, a long-period memory network module, a classifier and a decoder; the coding module is used for extracting the characteristics of each frame of image input into the forecast model by using the coder and outputting the extracted characteristics; the prediction module is used for carrying out sequence prediction on the output of the encoder by utilizing the long-short-period memory network module to obtain prediction characteristic information in the next period; the classification module is used for classifying the prediction characteristic information by using the classifier to obtain a first prediction forming state label; the decoding module is used for reconstructing the prediction characteristic information by utilizing the decoder to obtain a molten pool image in the next period; the image processing module is used for carrying out image processing on the molten pool image to obtain a second forecast forming state label; the fusion module is used for fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label; and the control module is used for controlling the additive manufacturing strategy of the next period according to the final forecast forming state label.
In general, through the above technical solutions conceived by the present invention, the following beneficial effects can be obtained: the model is suitable for prediction of the forming state of the additive manufacturing, the long-period memory network module is utilized for carrying out sequence prediction on the output of the encoder, and the reliability of the subsequent classification and the reconstructed image is improved, so that the accuracy of prediction control is improved; the long-period memory network is used as a sequence prediction, and another classifier is used independently, so that the classifier can set the output label types at will, and the label types output by the classifier can be set according to all possible situation types in the additive manufacturing forming process, so that various quality problems in the additive manufacturing process can be predicted accurately, and accurate control is realized; in addition, the video data in the additive manufacturing forming process is fully utilized, the forming state of the additive manufacturing can be accurately predicted based on continuous sampling images, and the accuracy of prediction control is further improved.
Drawings
FIG. 1 is a flow chart of a predictive control method for forming states of additive manufacturing according to an embodiment of the present invention;
fig. 2 is a block diagram of an additive manufacturing forming state prediction control system provided by an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention. In addition, the technical features of the embodiments of the present invention described below may be combined with each other as long as they do not collide with each other.
In the present invention, the terms "first," "second," and the like in the description and in the drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.
Fig. 1 is a flowchart of a method for predicting and controlling an forming state of additive manufacturing according to an embodiment of the present invention. Referring to fig. 1, the method for controlling the prediction of the forming state of the additive manufacturing in this embodiment includes operations S1-S4.
In this embodiment, the encoder-decoder neural network and the long-short-term memory neural network are used as trunks, and a neural network model of the encoder-long-term memory-classifier-decoder structure is built as a prediction model. The input of the forecast model is multi-frame continuous image data; the output of the prediction model comprises two parts, namely, each forming state label and probability output by the classifier and an image of a prediction forming area output by the encoder.
Before performing operation S1, the built prediction model needs to be trained, specifically including operation S0 '-operation S0' ".
And S0', carrying out additive manufacturing experiments by adopting different process parameters for a plurality of times to obtain various defects in each forming state, and collecting video data of an additive manufacturing forming area in the experimental process.
And S0' is operated, the video data is sampled to obtain an image set, the image set is divided into a plurality of groups of subsets by taking a preset length N and a preset step length as sliding windows, and x images after the subsets are marked to obtain a marked data set, wherein x is less than N/3.
Specifically, for example, a certain time length between 10ms and 200ms is selected as sampling time to sample video data, an ordered image set is obtained, all images in the image set are subjected to size transformation, the images are uniformly transformed into the input size of an encoder, and then segmentation and labeling operations are executed. The labeling is performed manually according to the forming state of the x images after each subset to obtain a labeling data set, and the manually labeled label comprises normal label, path error label, unstable welding machine process, welding wire shaking label, wire feeding disorder, welding bead flowing label, excessive slag label, shutdown label and the like.
Operation S0' "trains the predictive model with the top N-x images in each subset as input and the annotation dataset as output.
Further, the subsets obtained in operation S0 "are divided into training sets and test sets, and the dividing method includes, but is not limited to, leave-out method, cross-validation method, self-service method, and the like. In this embodiment, the subsets obtained in operation S0 "may be further expanded, where the expansion manners include, but are not limited to, random cropping, horizontal flipping, vertical flipping, mirroring, rotating different angles, adjusting brightness of a picture, adjusting contrast of a picture, adjusting chromaticity, changing the proportion of RGB color components, adjusting saturation of an image, performing gaussian blurring, sharpening, adding noise to an image, converting to a gray image, and the like. The expansion mode can be one or a plurality of expansion modes.
Preferably, an Adam optimization algorithm is utilized to train the forecasting model by taking the minimum loss function of the forecasting model as a training target, and the loss function adopts an average absolute error. After training, the structure and the weight of the forecast model are frozen and saved as a mode which can be called externally.
Operation S1, collecting video data of an additive manufacturing forming area and inputting the video data into a prediction model frame by frame, wherein the prediction model comprises an encoder, a long-term and short-term memory network module, a classifier and a decoder.
In the actual additive manufacturing control process, the additive manufacturing of the part is carried out according to initial technological parameters, and video data of a forming area and the vicinity thereof in the additive manufacturing process can be acquired in real time by using a set industrial camera and input into a trained forecasting model frame by frame.
And S2, performing feature extraction on each frame of image input into the prediction model by using the encoder, outputting the image, and performing sequence prediction on the output of the encoder by using the long-period memory network module to obtain prediction feature information in the next period.
The encoder performs feature extraction on each frame of image input into the prediction model through a single-layer convolutional neural network, a multi-layer convolutional neural network, a single-layer cyclic neural network or a multi-layer cyclic neural network and outputs the extracted image.
A Long Short-Term Memory (LSTM) network is a time-cycled neural network, has the characteristic of time Memory, and is suitable for processing and predicting important events with very Long intervals and delays in a time sequence. The LSTM module comprises single-layer or multi-layer LSTM and is used for realizing forecasting of time sequence characteristic information. In this embodiment, the long-short-term memory network module is used to perform sequence prediction on the output of the encoder, so as to obtain prediction feature information in the next time period, and improve the reliability of the subsequent classification and reconstructed image, thereby improving the accuracy of prediction control.
And S3, classifying the prediction characteristic information by using a classifier to obtain a first prediction forming state label, reconstructing the prediction characteristic information by using a decoder to obtain a molten pool image in the next period, and performing image processing on the molten pool image to obtain a second prediction forming state label.
In this embodiment, the long-short-term memory network is used as the sequence prediction, and another classifier is used alone, and because the classifier can set the output label type at will, the label type output by the classifier can be set according to all possible situation types in the additive manufacturing forming process, so as to accurately predict various quality problems in the additive manufacturing, and realize accurate control.
According to actual requirements, in the embodiment, the label categories output by the classifier include normal, path error, unstable welding machine process, welding wire shaking, wire feeding unsmooth, welding bead flowing, excessive slag and shutdown, the classifier also outputs the occurrence probability of each label category, and the sum of the occurrence probabilities of each label category is 1.
Based on this, in this embodiment, the first forecast forming state label and the second forecast forming state label each include a plurality of sub-labels and occurrence probabilities of the sub-labels. The plurality of sub-labels comprise normal, wrong paths, unstable welding machine process, welding wire shaking, unsmooth wire feeding, welding bead flowing, excessive slag and shutdown.
The decoder reconstructs the prediction characteristic information through a reverse network corresponding to the encoder. The decoder rebuilds the prediction characteristic information output by the long-period memory network module through a reverse network corresponding to the single-layer convolutional neural network, the multi-layer convolutional neural network, the single-layer convolutional neural network or the multi-layer convolutional neural network to obtain a molten pool image in the next period.
The second predictive modeling status tag may be obtained by image classification of the puddle image using a machine learning algorithm. Preferably, a k-Nearest Neighbor classification algorithm is selected, and a corresponding forming state probability vector is calculated based on the number of adjacent tag votes.
And S4, fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label, and controlling the additive manufacturing strategy of the next period according to the final forecast forming state label.
And carrying out fusion judgment on the first forecast forming state label and the second forecast forming state label by utilizing a decision fusion rule, respectively fusing the same sub-labels in the first forecast forming state label and the second forecast forming state label, and setting the sub-label with the highest occurrence probability after fusion as the final forecast forming state label. Specifically, the first forecast forming state label and the second forecast forming state label are subjected to average fusion, maximum fusion, weight fusion or generalized average fusion to obtain a final forecast forming state label. The decision fusion rule preferably adopts an average fusion and maximum support rule, the probability vector U1 in the first forecast forming state label and the probability vector U2 in the second forecast forming state label are fused and calculated, the probability vectors U3 and U3= (U1 + U2)/2 are obtained through fusion, and the sub label corresponding to the maximum probability value in the U3 is selected as the final forecast forming state label according to the maximum support rule.
The next period of additive manufacturing control strategy is, for example, process parameter modification, path modification, alarm or shutdown, etc. If the final forecast forming state label is normal, the additive manufacturing control strategy of the next period is to keep the original strategy unchanged so as to continue working according to the established technological parameters and technological paths. If the final forecast forming state label is a path error, the additive manufacturing control strategy of the next period is path modification so as to correspondingly adjust the process path in the additive manufacturing device, and the additive manufacturing device can continue the subsequent manufacturing after the adjustment is completed. If the final forecast forming state label is stop, the additive manufacturing control strategy of the next period is to stop the additive manufacturing device immediately so as to avoid material waste, product failure and serious defects. If the final forecast forming state label is unstable welding machine process, welding wire shaking, unsmooth wire feeding, welding bead flowing or excessive slag, the process parameters should be timely adjusted according to specific defect forms, and an additive manufacturing control strategy of the next period is formed. Therefore, when the defects do not appear, the method can forecast in advance, respond quickly, adjust the technological parameters in time or stop the machine, improve the stability and the quality of parts in the additive manufacturing process, and reduce the occurrence of waste products.
Fig. 2 is a block diagram of an additive manufacturing forming state prediction control system provided by an embodiment of the present invention. Referring to fig. 2, the additive manufacturing forming state prediction control system includes an acquisition module, a coding module, a prediction module, a classification module, a decoding module, an image processing module, a fusion module, and a control module.
The acquisition module is used for acquiring video data of the additive manufacturing forming area and inputting the video data into the forecasting model frame by frame, and the forecasting model comprises an encoder, a long-period memory network module, a classifier and a decoder. The coding module is used for extracting the characteristics of each frame of image input into the prediction model by using the coder and outputting the extracted characteristics. The prediction module is used for carrying out sequence prediction on the output of the encoder by utilizing the long-short-period memory network module to obtain prediction characteristic information in the next period. The classification module is used for classifying the prediction characteristic information by using a classifier to obtain a first prediction forming state label. The decoding module is used for reconstructing the prediction characteristic information by using the decoder to obtain a molten pool image in the next time period. The image processing module is used for carrying out image processing on the molten pool image to obtain a second forecast forming state label. The fusion module is used for fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label. The control module is used for controlling the additive manufacturing strategy of the next period according to the final forecast forming state label.
The additive manufacturing forming state prediction control system is configured to perform the additive manufacturing forming state prediction control method in the embodiment shown in fig. 1 described above. For details of this embodiment, please refer to the method for predicting and controlling the forming state of the additive manufacturing in the embodiment shown in fig. 1, which is not described herein.
It will be readily appreciated by those skilled in the art that the foregoing description is merely a preferred embodiment of the invention and is not intended to limit the invention, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (8)
1. A method for predictive control of an additive manufacturing forming state, comprising:
s1, collecting video data of an additive manufacturing forming area and inputting the video data into a prediction model frame by frame, wherein the prediction model comprises an encoder, a long-period and short-period memory network module, a classifier and a decoder;
s2, extracting features of each frame of image input into the prediction model by using the encoder, outputting the extracted features, and carrying out sequence prediction on the output of the encoder by using the long-period and short-period memory network module to obtain prediction feature information in the next period;
s3, classifying the prediction feature information by using the classifier to obtain a first prediction forming state label, reconstructing the prediction feature information by using the decoder to obtain a molten pool image in the next period, and performing image processing on the molten pool image to obtain a second prediction forming state label;
s4, fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label, and controlling an additive manufacturing strategy of the next period according to the final forecast forming state label;
the fusing of the first forecast forming state label and the second forecast forming state label in S4 includes: and respectively fusing the same sub-labels in the first forecast forming state label and the second forecast forming state label, and setting the sub-label with the highest occurrence probability after fusion as the final forecast forming state label.
2. The additive manufacturing forming state prediction control method according to claim 1, wherein the first predictive forming state label and the second predictive forming state label each include a plurality of sub-labels and occurrence probabilities of each of the sub-labels, the plurality of sub-labels including normal, path error, unstable welder process, wire shake, wire feed failure, bead flow, excessive slag, and shutdown.
3. The additive manufacturing forming state prediction control method according to claim 1 or 2, wherein the fusion in S4 is an average fusion, a maximum fusion, a weight fusion, or a generalized average fusion.
4. The additive manufacturing forming state prediction control method according to claim 1, characterized in that the S1 further includes, before:
carrying out additive manufacturing experiments by adopting different technological parameters for a plurality of times to obtain various defects in each forming state, and collecting video data of an additive manufacturing forming area in the experimental process;
sampling the video data to obtain an image set, dividing the image set into a plurality of groups of subsets by taking a preset length N and a preset step length as sliding windows, and labeling x images in each subset to obtain a labeling data set, wherein x is less than N/3;
the predictive model is trained with the top N-x images in each of the subsets as input and the annotation dataset as output.
5. The additive manufacturing forming state prediction control method of claim 4, wherein the training the predictive model comprises: and training the forecasting model by using a Adam optimization algorithm and taking the minimum loss function of the forecasting model as a training target, wherein the loss function adopts an average absolute error.
6. The method according to claim 1, wherein the encoder in S2 performs feature extraction on each frame of image input to the prediction model by using a single-layer convolutional neural network, a multi-layer convolutional neural network, a single-layer convolutional neural network, or a multi-layer convolutional neural network, and outputs the extracted image.
7. The additive manufacturing forming state prediction control method according to claim 1 or 6, characterized in that the decoder reconstructs the prediction characteristic information through a reverse network corresponding to the encoder in S3.
8. An additive manufacturing forming state predictive control system, comprising:
the system comprises an acquisition module, a prediction model and a prediction module, wherein the acquisition module is used for acquiring video data of an additive manufacturing forming area and inputting the video data into the prediction model frame by frame, and the prediction model comprises an encoder, a long-period memory network module, a classifier and a decoder;
the coding module is used for extracting the characteristics of each frame of image input into the forecast model by using the coder and outputting the extracted characteristics;
the prediction module is used for carrying out sequence prediction on the output of the encoder by utilizing the long-short-period memory network module to obtain prediction characteristic information in the next period;
the classification module is used for classifying the prediction characteristic information by using the classifier to obtain a first prediction forming state label;
the decoding module is used for reconstructing the prediction characteristic information by utilizing the decoder to obtain a molten pool image in the next period;
the image processing module is used for carrying out image processing on the molten pool image to obtain a second forecast forming state label;
the fusion module is used for fusing the first forecast forming state label and the second forecast forming state label to obtain a final forecast forming state label;
the fusing the first forecast forming state label and the second forecast forming state label in the fusing module comprises the following steps: respectively fusing the same sub-labels in the first forecast forming state label and the second forecast forming state label, and setting the sub-label with the highest occurrence probability after fusion as the final forecast forming state label;
and the control module is used for controlling the additive manufacturing strategy of the next period according to the final forecast forming state label.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110973673.5A CN113762366B (en) | 2021-08-24 | 2021-08-24 | Method and system for predictive control of forming state of additive manufacturing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110973673.5A CN113762366B (en) | 2021-08-24 | 2021-08-24 | Method and system for predictive control of forming state of additive manufacturing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113762366A CN113762366A (en) | 2021-12-07 |
CN113762366B true CN113762366B (en) | 2024-02-06 |
Family
ID=78790962
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110973673.5A Active CN113762366B (en) | 2021-08-24 | 2021-08-24 | Method and system for predictive control of forming state of additive manufacturing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113762366B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114619119B (en) * | 2022-03-29 | 2023-01-13 | 北京理工大学 | Electric arc vibration material disk makes on-line monitoring and real-time control system |
CN117984006B (en) * | 2024-04-03 | 2024-07-02 | 国网山东省电力公司潍坊供电公司 | Welding quality prediction method, device and medium based on welding infrared video generation |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3579148A1 (en) * | 2018-06-08 | 2019-12-11 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method and apparatus for training, classification model, mobile terminal, and readable storage medium |
CN112801091A (en) * | 2021-01-26 | 2021-05-14 | 汕头大学 | Additive manufacturing part forming quality monitoring and predicting method based on deep learning |
CN112989703A (en) * | 2021-04-02 | 2021-06-18 | 唐剑军 | Feature fusion metal additive manufacturing fusion state real-time prediction method based on deep learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11079745B2 (en) * | 2015-11-25 | 2021-08-03 | Lawrence Livermore National Security, Llc | Rapid closed-loop control based on machine learning |
-
2021
- 2021-08-24 CN CN202110973673.5A patent/CN113762366B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3579148A1 (en) * | 2018-06-08 | 2019-12-11 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method and apparatus for training, classification model, mobile terminal, and readable storage medium |
CN112801091A (en) * | 2021-01-26 | 2021-05-14 | 汕头大学 | Additive manufacturing part forming quality monitoring and predicting method based on deep learning |
CN112989703A (en) * | 2021-04-02 | 2021-06-18 | 唐剑军 | Feature fusion metal additive manufacturing fusion state real-time prediction method based on deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN113762366A (en) | 2021-12-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113762366B (en) | Method and system for predictive control of forming state of additive manufacturing | |
CN112381788B (en) | Part surface defect increment detection method based on double-branch matching network | |
US20210034044A1 (en) | Method for quality controling of target object and edge computing device | |
CN110648310B (en) | Weak supervision casting defect identification method based on attention mechanism | |
CN111126453A (en) | Fine-grained image classification method and system based on attention mechanism and cut filling | |
KR20200012334A (en) | Method for detecting defects in semiconductor device | |
CN115516500A (en) | Anomaly detection based on autoencoder and clustering | |
CN110795990A (en) | Gesture recognition method for underwater equipment | |
CN111931819A (en) | Machine fault prediction and classification method based on deep learning | |
CN111047532B (en) | Low-illumination video enhancement method based on 3D convolutional neural network | |
CN110688927A (en) | Video action detection method based on time sequence convolution modeling | |
CN114332008B (en) | Unsupervised defect detection and positioning method based on multi-level feature reconstruction | |
CN112801091A (en) | Additive manufacturing part forming quality monitoring and predicting method based on deep learning | |
CN111695435A (en) | Driver behavior identification method based on deep hybrid coding and decoding neural network | |
CN114913599A (en) | Video abnormal behavior detection method and system based on automatic encoder | |
CN114529894A (en) | Rapid scene text detection method fusing hole convolution | |
CN117540779A (en) | Lightweight metal surface defect detection method based on double-source knowledge distillation | |
JP2021033564A (en) | Image classification method, image classification device and image classification program | |
CN117197530A (en) | Insulator defect identification method based on improved YOLOv8 model and cosine annealing learning rate decay method | |
CN114772208B (en) | Non-contact belt tearing detection system and method based on image segmentation | |
CN116863404A (en) | Video abnormal event detection method for sensing local response | |
CN114037960B (en) | Method and system for identifying state of flap valve based on machine vision | |
CN115484456A (en) | Video anomaly prediction method and device based on semantic clustering | |
CN115424187B (en) | Auxiliary driving method for multi-angle camera collaborative importance ranking constraint | |
CN114943863B (en) | Method for identifying water gauge number of ship based on image restoration |
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 |