Despite the growing application of additive manufacturing (AM) in fabricating complex designs, most machines suffer from small working envelopes and slow processing speeds. One workaround to the problem of small throughput in AM is to partition the volume of a desired object and fabricate sub-volumes in parallel. Prior related work has focused on two problems. One is the geometric division problem, disregarding AM benefits and challenges in determining partitions. Others attempt to install multiple AM processing heads on the same machine, ensuring seamless bonding between deposited material from different heads while avoiding interference among them. A missed opportunity lies in deploying many independent machines simultaneously while considering benefits and limitations of AM. To that end, objects too large to be fabricated on one machine, are divided primarily into cubes that exploit benefits of AM. Specifically, the cubes are hollowed out in the direction of printing to reduce weight while avoiding the need for support structure, and depending on load conditions, packed in different orientations to mitigate material anisotropy.
{"title":"Parallelized Additive Manufacturing of Variably Partitioned Volumes for Large Scale 3D Printing With Localized Quality","authors":"Mahmoud Dinar","doi":"10.1115/detc2020-22496","DOIUrl":"https://doi.org/10.1115/detc2020-22496","url":null,"abstract":"\u0000 Despite the growing application of additive manufacturing (AM) in fabricating complex designs, most machines suffer from small working envelopes and slow processing speeds. One workaround to the problem of small throughput in AM is to partition the volume of a desired object and fabricate sub-volumes in parallel. Prior related work has focused on two problems. One is the geometric division problem, disregarding AM benefits and challenges in determining partitions. Others attempt to install multiple AM processing heads on the same machine, ensuring seamless bonding between deposited material from different heads while avoiding interference among them. A missed opportunity lies in deploying many independent machines simultaneously while considering benefits and limitations of AM. To that end, objects too large to be fabricated on one machine, are divided primarily into cubes that exploit benefits of AM. Specifically, the cubes are hollowed out in the direction of printing to reduce weight while avoiding the need for support structure, and depending on load conditions, packed in different orientations to mitigate material anisotropy.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"218 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123257124","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Melt pool size is a critical intermediate measure that reflects the outcome of a laser powder bed fusion process setting. Reliable melt pool predictions prior to builds can help users to evaluate potential part defects such as lack of fusion and over melting. This paper develops a layer-wise Neighboring-Effect Modeling (L-NBEM) method to predict melt pool size for 3D builds. The proposed method employs a feedforward neural network model with ten layer-wise and track-wise input variables. An experimental build using a spiral concentrating scan pattern with varying laser power was conducted on the Additive Manufacturing Metrology Testbed at the National Institute of Standards and Technology. Training and validation data were collected from 21 completed layers of the build, with 6,192,495 digital commands and 118,928 in-situ melt pool coaxial images. The L-NBEM model using the neural network approach demonstrates a better performance of average predictive error (12.12%) by leave-one-out cross-validation method, which is lower than the benchmark NBEM model (15.23%), and the traditional power-velocity model (19.41%).
{"title":"3D Build Melt Pool Predictive Modeling for Powder Bed Fusion Additive Manufacturing","authors":"Zhuo Yang, Yan Lu, H. Yeung, Sundar Kirshnamurty","doi":"10.1115/detc2020-22662","DOIUrl":"https://doi.org/10.1115/detc2020-22662","url":null,"abstract":"\u0000 Melt pool size is a critical intermediate measure that reflects the outcome of a laser powder bed fusion process setting. Reliable melt pool predictions prior to builds can help users to evaluate potential part defects such as lack of fusion and over melting. This paper develops a layer-wise Neighboring-Effect Modeling (L-NBEM) method to predict melt pool size for 3D builds. The proposed method employs a feedforward neural network model with ten layer-wise and track-wise input variables. An experimental build using a spiral concentrating scan pattern with varying laser power was conducted on the Additive Manufacturing Metrology Testbed at the National Institute of Standards and Technology. Training and validation data were collected from 21 completed layers of the build, with 6,192,495 digital commands and 118,928 in-situ melt pool coaxial images. The L-NBEM model using the neural network approach demonstrates a better performance of average predictive error (12.12%) by leave-one-out cross-validation method, which is lower than the benchmark NBEM model (15.23%), and the traditional power-velocity model (19.41%).","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"110 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115846656","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a new method to generate an adaptive spiral tool path for 3-axis CNC machining of the complex freeform surface directly from its representation in the form of the point cloud. The algorithm first constructs the uniform 2D circular mesh-grid to compute the Z (CL) points by applying the tool inverse offset method (IOM). Adaptive grid refinement is carried out iteratively until the surface form errors converge below the prescribed tolerance limits in both circumferential (forward) and radial (step) directions. Adaptive CL points are further refined to minimize the no. of tool lifts and generate an optimum sequence of machining regions. The optimized CL points are post-processed to generate the final CNC part programs in the ISO format. The part programs generated by our algorithm were extensively tested for various case studies using the commercial CNC simulator. The results were compared with those from the commercial CAM software. Our system was found to generate more efficient tool paths in terms of enhanced productivity, part quality, and reduced memory requirement.
{"title":"Spiral Tool Path Generation for CNC Machining Using Cloud of Points","authors":"M. Dhanda, A. Kukreja, S. S. Pande","doi":"10.1115/detc2020-22032","DOIUrl":"https://doi.org/10.1115/detc2020-22032","url":null,"abstract":"\u0000 This paper presents a new method to generate an adaptive spiral tool path for 3-axis CNC machining of the complex freeform surface directly from its representation in the form of the point cloud. The algorithm first constructs the uniform 2D circular mesh-grid to compute the Z (CL) points by applying the tool inverse offset method (IOM). Adaptive grid refinement is carried out iteratively until the surface form errors converge below the prescribed tolerance limits in both circumferential (forward) and radial (step) directions. Adaptive CL points are further refined to minimize the no. of tool lifts and generate an optimum sequence of machining regions. The optimized CL points are post-processed to generate the final CNC part programs in the ISO format. The part programs generated by our algorithm were extensively tested for various case studies using the commercial CNC simulator. The results were compared with those from the commercial CAM software. Our system was found to generate more efficient tool paths in terms of enhanced productivity, part quality, and reduced memory requirement.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121576170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Aditya U. Kulkarni, A. Salado, Christian Wernz, Peng Xu
Verification activities increase an engineering team’s confidence in its system design meeting system requirements, which in turn are derived from stakeholder needs. Conventional wisdom suggests that the system design should be verified frequently to minimize the cost of rework as the system design matures. However, this strategy is based more on experience of engineers than on a theoretical foundation. In this paper, we develop a belief-based model of verification of system design, using a single system requirement as an abstraction, to determine the conditions under which it is cost effective for an organization to verify frequently. We study the model for a broad set of growth rates in verification setup and rework costs. Our results show that verifying a system design frequently is not always an optimal verification strategy. Instead, it is only an optimal strategy when the costs of reworking a faulty design increase at a certain rate as the design matures.
{"title":"Is Verifying Frequently an Optimal Strategy? A Belief-Based Model of Verification","authors":"Aditya U. Kulkarni, A. Salado, Christian Wernz, Peng Xu","doi":"10.1115/detc2020-22582","DOIUrl":"https://doi.org/10.1115/detc2020-22582","url":null,"abstract":"\u0000 Verification activities increase an engineering team’s confidence in its system design meeting system requirements, which in turn are derived from stakeholder needs. Conventional wisdom suggests that the system design should be verified frequently to minimize the cost of rework as the system design matures. However, this strategy is based more on experience of engineers than on a theoretical foundation. In this paper, we develop a belief-based model of verification of system design, using a single system requirement as an abstraction, to determine the conditions under which it is cost effective for an organization to verify frequently. We study the model for a broad set of growth rates in verification setup and rework costs. Our results show that verifying a system design frequently is not always an optimal verification strategy. Instead, it is only an optimal strategy when the costs of reworking a faulty design increase at a certain rate as the design matures.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"186 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122513908","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Design and evaluation of exoskeletons is often a time consuming and costly process that involves prototyping, human testing, and multiple design iterations. For active exoskeletons, the primary challenge is to detect the wearer’s movement intent and provide potent assistance, which often requires sophisticated control algorithms. The goal of this study is to integrate human musculoskeletal models with robot modeling and control for virtual human-in-the-loop evaluation of exoskeleton design and control. We present potential strategies for assisting various human motions such as squatting, lifting, walking, and running. Several exoskeleton designs (for back, upper extremity, and lower extremity) and their control methods are evaluated with an integrated human-in-the-loop simulation approach to study their functionalities and biomechanical effects on the wearer’ musculoskeletal system. We hope this simulation paradigm can be utilized for virtual design and evaluation of exoskeletons and pave the way to build or optimize exoskeletons.
{"title":"Predictive Human-in-the-Loop Simulations for Assistive Exoskeletons","authors":"Xianlian Zhou","doi":"10.1115/detc2020-22668","DOIUrl":"https://doi.org/10.1115/detc2020-22668","url":null,"abstract":"\u0000 Design and evaluation of exoskeletons is often a time consuming and costly process that involves prototyping, human testing, and multiple design iterations. For active exoskeletons, the primary challenge is to detect the wearer’s movement intent and provide potent assistance, which often requires sophisticated control algorithms. The goal of this study is to integrate human musculoskeletal models with robot modeling and control for virtual human-in-the-loop evaluation of exoskeleton design and control. We present potential strategies for assisting various human motions such as squatting, lifting, walking, and running. Several exoskeleton designs (for back, upper extremity, and lower extremity) and their control methods are evaluated with an integrated human-in-the-loop simulation approach to study their functionalities and biomechanical effects on the wearer’ musculoskeletal system. We hope this simulation paradigm can be utilized for virtual design and evaluation of exoskeletons and pave the way to build or optimize exoskeletons.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"351 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115303016","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Powder bed fusion (PBF) has become a widely used additive manufacturing (AM) technology to produce metallic parts. Since the PBF process is driven by a moving heat source, consistency in part production, particularly when varying geometries, has proven difficult. Thermal field evolution during the manufacturing process determines both geometric and mechanical properties of the fabricated components. Simulations of the thermal field evolution can provide insight into desired process parameter selection for a given material and geometry. Thermal simulation of the PBF process is computationally challenging due to the geometric complexity of the manufacturing process and the inherent computational complexity that requires a numerical solution at every time increment of the process. We propose a new thermal simulation of the PBF process based on the laser scan path. Our approach is unique in that it does not restrict itself to simulations on the part design geometry, but instead simulates the formation of the geometry based on the process plan of a part. The implication of this distinction is that the simulations are in tune with the as-manufactured geometry, meaning that calculations are more aligned with the process than the design, and thus could be argued is a more realistic abstraction of real-world behavior. The discretization is based on the laser scan path, and the thermal model is formulated directly in terms of the manufacturing primitives. An element growth mechanism is introduced to simulate the evolution of a melt pool during the manufacturing process. A spatial data structure, called contact graph, is used to represent the discretized domain and capture all thermal interactions during the simulation. The simulation is localized through exploiting spatial and temporal locality, which is based on known empirical data. This limits the need to update to at most a constant number of elements at each time step. This implies that the proposed simulation not only scales to handle three-dimensional (3D) printed components of arbitrary complexity but also can achieve real-time performance. The simulation is fully implemented and validated against experimental data and other simulation results.
{"title":"Scalable Thermal Simulation of Powder Bed Fusion","authors":"Yaqi Zhang, V. Shapiro, P. Witherell","doi":"10.1115/detc2020-22628","DOIUrl":"https://doi.org/10.1115/detc2020-22628","url":null,"abstract":"\u0000 Powder bed fusion (PBF) has become a widely used additive manufacturing (AM) technology to produce metallic parts. Since the PBF process is driven by a moving heat source, consistency in part production, particularly when varying geometries, has proven difficult. Thermal field evolution during the manufacturing process determines both geometric and mechanical properties of the fabricated components. Simulations of the thermal field evolution can provide insight into desired process parameter selection for a given material and geometry. Thermal simulation of the PBF process is computationally challenging due to the geometric complexity of the manufacturing process and the inherent computational complexity that requires a numerical solution at every time increment of the process.\u0000 We propose a new thermal simulation of the PBF process based on the laser scan path. Our approach is unique in that it does not restrict itself to simulations on the part design geometry, but instead simulates the formation of the geometry based on the process plan of a part. The implication of this distinction is that the simulations are in tune with the as-manufactured geometry, meaning that calculations are more aligned with the process than the design, and thus could be argued is a more realistic abstraction of real-world behavior. The discretization is based on the laser scan path, and the thermal model is formulated directly in terms of the manufacturing primitives. An element growth mechanism is introduced to simulate the evolution of a melt pool during the manufacturing process. A spatial data structure, called contact graph, is used to represent the discretized domain and capture all thermal interactions during the simulation. The simulation is localized through exploiting spatial and temporal locality, which is based on known empirical data. This limits the need to update to at most a constant number of elements at each time step. This implies that the proposed simulation not only scales to handle three-dimensional (3D) printed components of arbitrary complexity but also can achieve real-time performance. The simulation is fully implemented and validated against experimental data and other simulation results.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127058850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Vishnu Aishwaryan Subra Mani, Nathaniel Goldfarb, G. Fischer
Over the past decade, wearable robotics and exoskeletons have been gaining recognition in the field of medical, assistive and augmentative robotics and have led to numerous new innovative mechanisms and designs. Due to fast-paced research activities, the critical importance and performance of established mechanisms such as wrap spring clutch/brake, Wafer Disc Brakes have been overlooked or used ineffectively. This paper describes a practical design approach that will enable the designer to choose a mechanism based on the application of the device, which will promote overall growth in current technology. The Legged Anthropomorphic Robotic Rehabilitation Exoskeleton (LARRE) project used this approach to design, manufacture, and test the knee joint for ground-level walking. This paper provides the reasoning behind the selection of wrap spring clutch, its evaluation, and testing standards as the knee joint. A thorough literature review was conducted to understand the current state of the art. This project collected a rich set of biomechanical data to ensure that the mechanism will produce the right moments and range of motions during walking. To ensure that our mechanism meets the requirements, the mechanism was put through a wide range of stress tests. The paper establishes a methodology to choose a mechanism for an exoskeleton’s joint based on the desired requirements. The outcome of this paper is an analytical based design approach that can be used by other researchers to impart additional traits and weights, which will aid in the development of exoskeleton design.
{"title":"Design, Development and Characterization of a Wrap Spring Clutch/Brake Mechanism As a Knee Joint for an Assistive Exoskeleton","authors":"Vishnu Aishwaryan Subra Mani, Nathaniel Goldfarb, G. Fischer","doi":"10.1115/detc2020-22444","DOIUrl":"https://doi.org/10.1115/detc2020-22444","url":null,"abstract":"\u0000 Over the past decade, wearable robotics and exoskeletons have been gaining recognition in the field of medical, assistive and augmentative robotics and have led to numerous new innovative mechanisms and designs. Due to fast-paced research activities, the critical importance and performance of established mechanisms such as wrap spring clutch/brake, Wafer Disc Brakes have been overlooked or used ineffectively. This paper describes a practical design approach that will enable the designer to choose a mechanism based on the application of the device, which will promote overall growth in current technology. The Legged Anthropomorphic Robotic Rehabilitation Exoskeleton (LARRE) project used this approach to design, manufacture, and test the knee joint for ground-level walking. This paper provides the reasoning behind the selection of wrap spring clutch, its evaluation, and testing standards as the knee joint. A thorough literature review was conducted to understand the current state of the art. This project collected a rich set of biomechanical data to ensure that the mechanism will produce the right moments and range of motions during walking. To ensure that our mechanism meets the requirements, the mechanism was put through a wide range of stress tests. The paper establishes a methodology to choose a mechanism for an exoskeleton’s joint based on the desired requirements. The outcome of this paper is an analytical based design approach that can be used by other researchers to impart additional traits and weights, which will aid in the development of exoskeleton design.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122210581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
M. Rossoni, P. Bolzan, G. Colombo, M. Bordegoni, M. Carulli
During the concept phase of the industrial design process drawings are used to represent designer’s ideas. More specifically, the designer’s goal is to put the characteristics of ideas on paper so that they can later act as pivotal points in the development of a project. Sketching is also the ideal tool to continue developing an idea: because it is imprecise, the sketch guarantees a high degree of freedom, allowing for changes to made and new ideas to be added. Another possibility is to translate ideas into sketches on computer tools. This approach can allow the designer to use the created 3D model as the basis for further developing ideas. At the present moment, however, this type of solution is not extensively used by designers during the concept phase. Some researchers have identified technical problems as the reason why these instruments have been unsuccessful on the market, while for others this is related to systems still too rigid to be adapted to the often-diverse needs of designers. The research presented in this position paper aims at analyzing what has so far been understood with respect to the process of generating ideas, their initial representation in the concept phase and the tools that have been developed so far to support this phase. Consequently, a discussion on these themes and some hypotheses from which develop new research lines will be presented.
{"title":"Survey of Digital Tools for the Generation of Ideas","authors":"M. Rossoni, P. Bolzan, G. Colombo, M. Bordegoni, M. Carulli","doi":"10.1115/detc2020-22443","DOIUrl":"https://doi.org/10.1115/detc2020-22443","url":null,"abstract":"\u0000 During the concept phase of the industrial design process drawings are used to represent designer’s ideas. More specifically, the designer’s goal is to put the characteristics of ideas on paper so that they can later act as pivotal points in the development of a project. Sketching is also the ideal tool to continue developing an idea: because it is imprecise, the sketch guarantees a high degree of freedom, allowing for changes to made and new ideas to be added.\u0000 Another possibility is to translate ideas into sketches on computer tools. This approach can allow the designer to use the created 3D model as the basis for further developing ideas. At the present moment, however, this type of solution is not extensively used by designers during the concept phase. Some researchers have identified technical problems as the reason why these instruments have been unsuccessful on the market, while for others this is related to systems still too rigid to be adapted to the often-diverse needs of designers.\u0000 The research presented in this position paper aims at analyzing what has so far been understood with respect to the process of generating ideas, their initial representation in the concept phase and the tools that have been developed so far to support this phase. Consequently, a discussion on these themes and some hypotheses from which develop new research lines will be presented.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131312836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
One of the most prevalent additive manufacturing processes, the powder bed fusion process, is driven by a moving heat source that melts metals to build a part. This moving heat source, and the subsequent formation and moving of a melt pool, plays an important role in determining both the geometric and mechanical properties of the printed components. The ability to control the melt pool during the build process is a sought after mechanism for improving quality control and optimizing manufacturing parameters. For this reason, efficient models that can predict melt pool size based on the process input (i.e., laser power, scan speed, spot size and scan path) offer a path to improved process control. Towards improved process control, a data-driven melt pool prediction model is built with a neighborhood-based neural network and trained using experimental data from the National Institute of Standards and Technology (NIST). The model considers the influence of both manufacturing parameters and laser scan paths. The scan path information is encoded using two novel neighborhood features of the neural network through locality. The neural network is used to generate a surrogate model, and we demonstrate how the performance of the resulting surrogate model can be further improved by using several ensemble methods. We then demonstrate how the trained surrogate model can be used as a forward solver for developing novel laser power design algorithms. The resulting laser power plan is designed to keep melt pool size as constant as possible for any given scan pattern. The algorithm is implemented and validated with numerical experiments.
{"title":"A Neighborhood-Based Neural Network for Melt Pool Prediction and Control","authors":"Yaqi Zhang, V. Shapiro, P. Witherell","doi":"10.1115/detc2020-22549","DOIUrl":"https://doi.org/10.1115/detc2020-22549","url":null,"abstract":"\u0000 One of the most prevalent additive manufacturing processes, the powder bed fusion process, is driven by a moving heat source that melts metals to build a part. This moving heat source, and the subsequent formation and moving of a melt pool, plays an important role in determining both the geometric and mechanical properties of the printed components. The ability to control the melt pool during the build process is a sought after mechanism for improving quality control and optimizing manufacturing parameters. For this reason, efficient models that can predict melt pool size based on the process input (i.e., laser power, scan speed, spot size and scan path) offer a path to improved process control.\u0000 Towards improved process control, a data-driven melt pool prediction model is built with a neighborhood-based neural network and trained using experimental data from the National Institute of Standards and Technology (NIST). The model considers the influence of both manufacturing parameters and laser scan paths. The scan path information is encoded using two novel neighborhood features of the neural network through locality. The neural network is used to generate a surrogate model, and we demonstrate how the performance of the resulting surrogate model can be further improved by using several ensemble methods. We then demonstrate how the trained surrogate model can be used as a forward solver for developing novel laser power design algorithms. The resulting laser power plan is designed to keep melt pool size as constant as possible for any given scan pattern. The algorithm is implemented and validated with numerical experiments.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133626937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
An experienced engineer can glance at a component and suggest appropriate methods for its manufacture. This skill has been difficult to automate but in recent years Neural Networks have demonstrated impressive image recognition capabilities in many applications. Consequently, this work is motivated by the goal of automating shape assessment for manufacturing. Specifically the reported work investigates the feasibility of training a convolutional neural network (CNN) to recognize 2D images of shapes associated with particular Near Net Shape (NNS) manufacturing processes such as casting, forging, or flow forming. The system uses multiple images generated from 3D CAD models (each manually associated with specific NNS processes) as training data and a single shop floor photograph as a classification query. While multiple views are used to train the CNN only a single view is used to assess the accuracy of the classification. Such single-view classification is designed to support the easy assessment of physical parts observed in manufacturing facilities where it would often be impractical to create an array of images from many viewpoints. The result suggests that despite limitations, single-view CNNs can classify real engineering components for manufacture.
{"title":"Automated Classification of Components for Manufacturing Planning: Single-View Convolutional Neural Network for Global Shape Identification","authors":"Andy Barclay, J. Corney","doi":"10.1115/detc2020-22335","DOIUrl":"https://doi.org/10.1115/detc2020-22335","url":null,"abstract":"\u0000 An experienced engineer can glance at a component and suggest appropriate methods for its manufacture. This skill has been difficult to automate but in recent years Neural Networks have demonstrated impressive image recognition capabilities in many applications. Consequently, this work is motivated by the goal of automating shape assessment for manufacturing. Specifically the reported work investigates the feasibility of training a convolutional neural network (CNN) to recognize 2D images of shapes associated with particular Near Net Shape (NNS) manufacturing processes such as casting, forging, or flow forming. The system uses multiple images generated from 3D CAD models (each manually associated with specific NNS processes) as training data and a single shop floor photograph as a classification query. While multiple views are used to train the CNN only a single view is used to assess the accuracy of the classification. Such single-view classification is designed to support the easy assessment of physical parts observed in manufacturing facilities where it would often be impractical to create an array of images from many viewpoints. The result suggests that despite limitations, single-view CNNs can classify real engineering components for manufacture.","PeriodicalId":164403,"journal":{"name":"Volume 9: 40th Computers and Information in Engineering Conference (CIE)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129050432","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","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：