Digital maskless lithography is growing in popularity due to its unique ability to fabricate high-resolution parts at a fast speed without the need for physical masks. Though the theoretical foundation for photopolymerization exists, it is difficult to observe the voxel growth process in situ. This can be attributed to the low refractive index difference between cured and uncured resin, the microscopic size of the parts, and the rapid rate of photopolymerization after crossing the threshold. Therefore, a system that can address these issues is highly desired. Schlieren optics is a tool that makes the minute changes in the refractive indices visible. This paper proposes a modified schlieren-based observation system with confocal magnifying optics that create a virtual screen at the focal plane of the camera. The proposed technique visualizes the light deflection by the changing density induced refractive index gradient, and the use of focusing optics enables flexible positioning of the virtual screen and optical magnification. Single-shot binary images with a different number of pixels were used for fabricating voxels. Different factors affecting the voxel shape like chemical composition, energy input is studied. The observed results are compared against simulations based on Beer-Lambert's law, photopolymerization curve, and Gaussian beam propagation theory. The physical experimental results demonstrated the effectiveness of the proposed observation system. Application of this system in fabrication of microlenses and its advantages over theoretical model-based profile predictions are briefly discussed.
{"title":"Enhanced Schlieren System for In-Situ Observation of Dynamic Light-Resin Interactions in Projection-based Stereolithography Process","authors":"Aditya Chivate, Chi Zhou","doi":"10.1115/1.4062218","DOIUrl":"https://doi.org/10.1115/1.4062218","url":null,"abstract":"\u0000 Digital maskless lithography is growing in popularity due to its unique ability to fabricate high-resolution parts at a fast speed without the need for physical masks. Though the theoretical foundation for photopolymerization exists, it is difficult to observe the voxel growth process in situ. This can be attributed to the low refractive index difference between cured and uncured resin, the microscopic size of the parts, and the rapid rate of photopolymerization after crossing the threshold. Therefore, a system that can address these issues is highly desired. Schlieren optics is a tool that makes the minute changes in the refractive indices visible. This paper proposes a modified schlieren-based observation system with confocal magnifying optics that create a virtual screen at the focal plane of the camera. The proposed technique visualizes the light deflection by the changing density induced refractive index gradient, and the use of focusing optics enables flexible positioning of the virtual screen and optical magnification. Single-shot binary images with a different number of pixels were used for fabricating voxels. Different factors affecting the voxel shape like chemical composition, energy input is studied. The observed results are compared against simulations based on Beer-Lambert's law, photopolymerization curve, and Gaussian beam propagation theory. The physical experimental results demonstrated the effectiveness of the proposed observation system. Application of this system in fabrication of microlenses and its advantages over theoretical model-based profile predictions are briefly discussed.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48286766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yaoyao Ping, Yongkui Liu, Lin Zhang, Lihui Wang, Xun Xu
Cloud manufacturing is a manufacturing model that aims to provide on-demand resources and services to consumers over the Internet. Scheduling is one of the core techniques for cloud manufacturing to achieve the aim. Multi-task scheduling with dynamical task arrivals is an important research issue in the area of cloud manufacturing scheduling. Many traditional algorithms such as the genetic algorithm (GA) and ant colony optimization algorithm (ACO) have been used to solve the issue, which, however, are either incapable of or perform poorly in tackling the problem. Deep reinforcement learning (DRL) that combines artificial neural networks with reinforcement learning provides an effective technique in this regard. In view of this, we employ a typical deep reinforcement learning algorithm – Deep Q-network (DQN) – and proposed a DQN-based multi-task scheduling approach for cloud manufacturing. Three different task arrival modes – arriving at the same time, arriving in random batches, and arriving one by one sequentially – are considered. Four baseline approaches including random scheduling, round-robin scheduling, earliest scheduling, and minimum execution time scheduling are investigated. A comparison of results indicates that the DQN-based scheduling approach is able to effectively address the multi-task scheduling problem in cloud manufacturing and performs best among all approaches.
{"title":"Deep Reinforcement Learning-Based Multi-Task Scheduling in Cloud Manufacturing under Different Task Arrival Modes","authors":"Yaoyao Ping, Yongkui Liu, Lin Zhang, Lihui Wang, Xun Xu","doi":"10.1115/1.4062217","DOIUrl":"https://doi.org/10.1115/1.4062217","url":null,"abstract":"\u0000 Cloud manufacturing is a manufacturing model that aims to provide on-demand resources and services to consumers over the Internet. Scheduling is one of the core techniques for cloud manufacturing to achieve the aim. Multi-task scheduling with dynamical task arrivals is an important research issue in the area of cloud manufacturing scheduling. Many traditional algorithms such as the genetic algorithm (GA) and ant colony optimization algorithm (ACO) have been used to solve the issue, which, however, are either incapable of or perform poorly in tackling the problem. Deep reinforcement learning (DRL) that combines artificial neural networks with reinforcement learning provides an effective technique in this regard. In view of this, we employ a typical deep reinforcement learning algorithm – Deep Q-network (DQN) – and proposed a DQN-based multi-task scheduling approach for cloud manufacturing. Three different task arrival modes – arriving at the same time, arriving in random batches, and arriving one by one sequentially – are considered. Four baseline approaches including random scheduling, round-robin scheduling, earliest scheduling, and minimum execution time scheduling are investigated. A comparison of results indicates that the DQN-based scheduling approach is able to effectively address the multi-task scheduling problem in cloud manufacturing and performs best among all approaches.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45880001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this work, we developed a new additive manufacturing paradigm, coaxial wire-powder fed directed energy deposition (CWP-DED), to enable the fabrication of metals or composites with high manufacturing flexibility and efficiency. Herein, stainless steel (SS) 316L was selected as a representative material to validate the feasibility of CWP-DED process. Effects of feed rates on the melt pool thermodynamics during the CWP-DED process were investigated using experimental and analytical approaches. Thermal contributions of fed wire and powders to the melt pool were involved in the analytical model to predict the melt pool temperature. The experimental results from thermal imaging were also obtained for validation. Besides, we uncovered the evolution of solidification morphology and crystallographic texture with different combinations of wire and powder feed rates. Finally, the microhardness and tensile performance of different as-built parts were tested. The results showed that the powder feed rate played a more dominant role in determining the melt pool temperature than the wire feed rate. Melt pool temperature experienced an initial increase and then decrease with the powder feed rate. A fine microstructure was achieved at a low powder feed rate, producing higher microhardness and larger tensile strength. This paper revealed the relations among process, thermal variation, microstructures, and mechanical properties of as-built metallic parts to provide a fundamental understanding of this novel DED process.
{"title":"Directed Energy Deposition with Coaxial Wire-Powder Feeding: Melt Pool Temperature and Microstructure","authors":"Yue Zhou, F. Ning","doi":"10.1115/1.4062216","DOIUrl":"https://doi.org/10.1115/1.4062216","url":null,"abstract":"\u0000 In this work, we developed a new additive manufacturing paradigm, coaxial wire-powder fed directed energy deposition (CWP-DED), to enable the fabrication of metals or composites with high manufacturing flexibility and efficiency. Herein, stainless steel (SS) 316L was selected as a representative material to validate the feasibility of CWP-DED process. Effects of feed rates on the melt pool thermodynamics during the CWP-DED process were investigated using experimental and analytical approaches. Thermal contributions of fed wire and powders to the melt pool were involved in the analytical model to predict the melt pool temperature. The experimental results from thermal imaging were also obtained for validation. Besides, we uncovered the evolution of solidification morphology and crystallographic texture with different combinations of wire and powder feed rates. Finally, the microhardness and tensile performance of different as-built parts were tested. The results showed that the powder feed rate played a more dominant role in determining the melt pool temperature than the wire feed rate. Melt pool temperature experienced an initial increase and then decrease with the powder feed rate. A fine microstructure was achieved at a low powder feed rate, producing higher microhardness and larger tensile strength. This paper revealed the relations among process, thermal variation, microstructures, and mechanical properties of as-built metallic parts to provide a fundamental understanding of this novel DED process.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":"60 3","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"41315320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Cutting force identification is critical to improving industrial robot performance and reducing machining vibration. However, most indirect identification methods of cutting force are not applicable since the dynamic characteristics of the robotic milling system vary with the robot pose. In this paper, a novel pose-dependent method is proposed to identify the cutting force using the acceleration signal generated by robotic milling. Firstly, the modal parameters of the robot at different machining points are used as a training dataset to develop the Gaussian Process Regression (GPR) model. Next, the modal parameters predicted by the GPR model are used to optimize the cutting force estimation based on the minimum variance unbiased estimate method. Then, the Kalman filter method is used to update the covariance matrix of the cutting force identification error and the state estimation error. Lastly, the proposed method is verified with the experiment, and the results show that the identification error and time are acceptable under the condition of variable robot pose.
{"title":"Pose-dependent Cutting Force Identification for Robotic Milling","authors":"Maxiao Hou, Hongru Cao, Yang Luo, Yanjie Guo","doi":"10.1115/1.4062145","DOIUrl":"https://doi.org/10.1115/1.4062145","url":null,"abstract":"\u0000 Cutting force identification is critical to improving industrial robot performance and reducing machining vibration. However, most indirect identification methods of cutting force are not applicable since the dynamic characteristics of the robotic milling system vary with the robot pose. In this paper, a novel pose-dependent method is proposed to identify the cutting force using the acceleration signal generated by robotic milling. Firstly, the modal parameters of the robot at different machining points are used as a training dataset to develop the Gaussian Process Regression (GPR) model. Next, the modal parameters predicted by the GPR model are used to optimize the cutting force estimation based on the minimum variance unbiased estimate method. Then, the Kalman filter method is used to update the covariance matrix of the cutting force identification error and the state estimation error. Lastly, the proposed method is verified with the experiment, and the results show that the identification error and time are acceptable under the condition of variable robot pose.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44107443","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chatter is one of the major issues that cause undesirable effects limiting machining productivity. Passive control devices, such as tuned mass dampers (TMDs), have been widely employed to increase machining stability by suppressing chatter. More recently, inerter-based devices have been developed for a wide variety of engineering vibration mitigation applications. However, no experimental study for the application of inerters to the machining stability problem has yet been conducted. This paper presents an implementation of an inerter-based dynamic vibration absorber (IDVA) to the problem of chatter stability, for the first time. For this, it employs the IDVA with a pivoted-bar inerter developed in [1] to mitigate the chatter effect under cutting forces in milling. Due to the nature of machining stability, the optimal design parameters for the IDVA are numerically obtained by considering the real part of the frequency response function (FRF) which enables the absolute stability limit in a single degree-of-freedom (SDOF) to be maximised for a milling operation. Chatter performance is experimentally validated through milling trials using the prototype IDVA and a flexible workpiece. The experimental results show that the IDVA provides more than 15% improvement in the absolute stability limit compared to a classical TMD.
{"title":"Implementation of inerter-based dynamic vibration absorber for chatter suppression","authors":"H. Dogan, N. Sims, D. Wagg","doi":"10.1115/1.4062118","DOIUrl":"https://doi.org/10.1115/1.4062118","url":null,"abstract":"\u0000 Chatter is one of the major issues that cause undesirable effects limiting machining productivity. Passive control devices, such as tuned mass dampers (TMDs), have been widely employed to increase machining stability by suppressing chatter. More recently, inerter-based devices have been developed for a wide variety of engineering vibration mitigation applications. However, no experimental study for the application of inerters to the machining stability problem has yet been conducted. This paper presents an implementation of an inerter-based dynamic vibration absorber (IDVA) to the problem of chatter stability, for the first time. For this, it employs the IDVA with a pivoted-bar inerter developed in [1] to mitigate the chatter effect under cutting forces in milling. Due to the nature of machining stability, the optimal design parameters for the IDVA are numerically obtained by considering the real part of the frequency response function (FRF) which enables the absolute stability limit in a single degree-of-freedom (SDOF) to be maximised for a milling operation. Chatter performance is experimentally validated through milling trials using the prototype IDVA and a flexible workpiece. The experimental results show that the IDVA provides more than 15% improvement in the absolute stability limit compared to a classical TMD.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44177303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. Zare, M. Tunesi, T. Harriman, John R. Troutman, M. Davies, D. Lucca
Single crystal Ge is a semiconductor that has broad applications, especially in manipulation of infra-red (IR) light. Diamond machining enables the efficient production of surfaces with tolerances required by the optical industry. During machining of anisotropic single crystals, the cutting direction with respect to the in-plane lattice orientation plays a fundamental role in the final quality of the surface and subsurface. In this study, on-axis face turning experiments were performed on an undoped (111)Ge wafer to investigate the effects of crystal anisotropy and feedrate on the surface and subsurface condition. Atomic force microscopy and scanning white light interferometry were used to characterize the presence of brittle fracture on the machined surfaces and to evaluate the resultant surface roughness. Raman spectroscopy was performed to evaluate the residual stresses and lattice disorder induced by the tool during machining. Nanoindentation with Berkovich and cube corner indenter tips was performed to evaluate elastic modulus, hardness, and fracture toughness of the machined surfaces and to study their variations with feedrate and cutting direction. Post-indentation studies of selected indentations were also performed to characterize the corresponding quasi-plasticity mechanisms. It was found that an increase of feedrate produced a rotation of the resultant force imparted by the tool indication a shift from indentation-dominant to cutting-dominant behavior. Fracture increased with the feedrate and showed a higher propensity when the cutting direction belonged to the <112¯> family.
{"title":"Face Turning of Single Crystal (111)Ge: Cutting Mechanics and Surface/Subsurface Characteristics","authors":"A. Zare, M. Tunesi, T. Harriman, John R. Troutman, M. Davies, D. Lucca","doi":"10.1115/1.4057054","DOIUrl":"https://doi.org/10.1115/1.4057054","url":null,"abstract":"\u0000 Single crystal Ge is a semiconductor that has broad applications, especially in manipulation of infra-red (IR) light. Diamond machining enables the efficient production of surfaces with tolerances required by the optical industry. During machining of anisotropic single crystals, the cutting direction with respect to the in-plane lattice orientation plays a fundamental role in the final quality of the surface and subsurface. In this study, on-axis face turning experiments were performed on an undoped (111)Ge wafer to investigate the effects of crystal anisotropy and feedrate on the surface and subsurface condition. Atomic force microscopy and scanning white light interferometry were used to characterize the presence of brittle fracture on the machined surfaces and to evaluate the resultant surface roughness. Raman spectroscopy was performed to evaluate the residual stresses and lattice disorder induced by the tool during machining. Nanoindentation with Berkovich and cube corner indenter tips was performed to evaluate elastic modulus, hardness, and fracture toughness of the machined surfaces and to study their variations with feedrate and cutting direction. Post-indentation studies of selected indentations were also performed to characterize the corresponding quasi-plasticity mechanisms. It was found that an increase of feedrate produced a rotation of the resultant force imparted by the tool indication a shift from indentation-dominant to cutting-dominant behavior. Fracture increased with the feedrate and showed a higher propensity when the cutting direction belonged to the <112¯> family.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42249345","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Editorial","authors":"Albert Shih, Ajay P. Malshe","doi":"10.1115/1.4057045","DOIUrl":"https://doi.org/10.1115/1.4057045","url":null,"abstract":"\u0000 May 2023 Editorial","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42825804","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Bo Pan, R. Kang, Xu Zhu, Zhe Yang, Juntao Zhang, Jiang Guo
Double-sided lapping (DSL) is a precision process widely used for machining flat workpieces, such as optical windows, wafers, and brake pads owing to its high efficiency and parallelism. However, the mechanism of parallelism error reduced by the DSL process was rarely investigated. Furthermore, the relationship between parallelism and the flatness was not clearly illustrated. To explain why the parallelism of workpieces becomes convergent by the DSL, a theoretical model has been developed in this paper by calculating the parallelism evolution with the consideration of variation contact situations between workpieces and lapping plates for the first time. Moreover, several workpieces, including a slanted one rendering the model close to the actual process, are taken to calculate the parallelism evolution, and the mechanism of the parallelism error reduced by the DSL process is clarified. The calculation result has indicated that the parallelism error was reduced from 100.0 μm to 25.6 μm based on the parallelism evolution model. The experimental results showed that the parallelism improved from 108.6 μm to 28.2 μm, which agreed with the theoretical results well.
{"title":"Why parallelism of workpieces becomes convergent during double-sided lapping?","authors":"Bo Pan, R. Kang, Xu Zhu, Zhe Yang, Juntao Zhang, Jiang Guo","doi":"10.1115/1.4057053","DOIUrl":"https://doi.org/10.1115/1.4057053","url":null,"abstract":"\u0000 Double-sided lapping (DSL) is a precision process widely used for machining flat workpieces, such as optical windows, wafers, and brake pads owing to its high efficiency and parallelism. However, the mechanism of parallelism error reduced by the DSL process was rarely investigated. Furthermore, the relationship between parallelism and the flatness was not clearly illustrated. To explain why the parallelism of workpieces becomes convergent by the DSL, a theoretical model has been developed in this paper by calculating the parallelism evolution with the consideration of variation contact situations between workpieces and lapping plates for the first time. Moreover, several workpieces, including a slanted one rendering the model close to the actual process, are taken to calculate the parallelism evolution, and the mechanism of the parallelism error reduced by the DSL process is clarified. The calculation result has indicated that the parallelism error was reduced from 100.0 μm to 25.6 μm based on the parallelism evolution model. The experimental results showed that the parallelism improved from 108.6 μm to 28.2 μm, which agreed with the theoretical results well.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42053092","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Industrial robots have become a suitable alternative to machine tools due to their great flexibility, low cost, and large working space. However, the deformation and vibration caused by the cutting forces during machining result in poor machining accuracy and surface quality. In order to improve the machining performance of the robot, this paper proposes a posture optimization method for robotic milling with the redundant degree of freedom of the industrial robot. First, modal tests are conducted in the robotic workspace to obtain the parameters of the structural dynamics of the robotic milling system. Then, considering the dynamics model of the system, the optimization model based on surface location error (SLE) is proposed to obtain the optimal robotic posture. Finally, a series of experiments illustrate that pose optimization based on SLE can improve the machining accuracy and surface machining quality.
{"title":"Pose optimization in robotic milling based on surface location error","authors":"Teng-fei Hou, Yang Lei, Ye Ding","doi":"10.1115/1.4057055","DOIUrl":"https://doi.org/10.1115/1.4057055","url":null,"abstract":"\u0000 Industrial robots have become a suitable alternative to machine tools due to their great flexibility, low cost, and large working space. However, the deformation and vibration caused by the cutting forces during machining result in poor machining accuracy and surface quality. In order to improve the machining performance of the robot, this paper proposes a posture optimization method for robotic milling with the redundant degree of freedom of the industrial robot. First, modal tests are conducted in the robotic workspace to obtain the parameters of the structural dynamics of the robotic milling system. Then, considering the dynamics model of the system, the optimization model based on surface location error (SLE) is proposed to obtain the optimal robotic posture. Finally, a series of experiments illustrate that pose optimization based on SLE can improve the machining accuracy and surface machining quality.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44879669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ayantha Senanayaka, Wenmeng Tian, T. Falls, L. Bian
This study aims to develop an intelligent, rapid porosity prediction methodology for varying process conditions based on knowledge transfer from the existing process conditions. Conventional machine learning algorithms are extensively used in porosity prediction. However, these approaches assume that the training (source) and testing (target) data follow the same probability distribution, and the labeled data are available in both source and target domains. The source and target do not follow the same distribution in real-world manufacturing environments. The diversity of industrialization processes leads to heterogeneous data collection in different production conditions, and labeling is costly. Transfer learning is one of the robust techniques that enables transferring learned knowledge between source and target to establish a relationship while the target has less data. Therefore, this paper presents similarity-based multi-source transfer learning(SiMuS-TL) method to develop a relationship between a source and an unknown target. The similarities between sources and targets are learned by forming a new domain called the mixed domain, which organizes data into identity groups. Then, a group-based learning process is designated to transfer knowledge to make target predictions. The effectiveness of the SiMuS-TL is explored with the application of porosity prediction in additively manufactured parts in realistic situations, i.e., single-source and multi-sources transfer to unknown target porosity prediction. The porosity prediction accuracies are approximately 90% for both scenarios with the SiMuS-TL method, but conventional SVM and CNN classifiers barely perform well in predicting porosity while process condition varies.
{"title":"Understanding the Effects of Process Conditions on Thermal-Defect Relationship: A Transfer Machine Learning Approach","authors":"Ayantha Senanayaka, Wenmeng Tian, T. Falls, L. Bian","doi":"10.1115/1.4057052","DOIUrl":"https://doi.org/10.1115/1.4057052","url":null,"abstract":"\u0000 This study aims to develop an intelligent, rapid porosity prediction methodology for varying process conditions based on knowledge transfer from the existing process conditions. Conventional machine learning algorithms are extensively used in porosity prediction. However, these approaches assume that the training (source) and testing (target) data follow the same probability distribution, and the labeled data are available in both source and target domains. The source and target do not follow the same distribution in real-world manufacturing environments. The diversity of industrialization processes leads to heterogeneous data collection in different production conditions, and labeling is costly. Transfer learning is one of the robust techniques that enables transferring learned knowledge between source and target to establish a relationship while the target has less data. Therefore, this paper presents similarity-based multi-source transfer learning(SiMuS-TL) method to develop a relationship between a source and an unknown target. The similarities between sources and targets are learned by forming a new domain called the mixed domain, which organizes data into identity groups. Then, a group-based learning process is designated to transfer knowledge to make target predictions. The effectiveness of the SiMuS-TL is explored with the application of porosity prediction in additively manufactured parts in realistic situations, i.e., single-source and multi-sources transfer to unknown target porosity prediction. The porosity prediction accuracies are approximately 90% for both scenarios with the SiMuS-TL method, but conventional SVM and CNN classifiers barely perform well in predicting porosity while process condition varies.","PeriodicalId":16299,"journal":{"name":"Journal of Manufacturing Science and Engineering-transactions of The Asme","volume":"25 12","pages":""},"PeriodicalIF":4.0,"publicationDate":"2023-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"41297803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：