The issue of training operators in the use of machinery is topical in the industrial field and in many other contexts, such as university laboratories. Training is about learning how to use machinery properly and safely. Beyond the possibility of studying manuals to learn how to use a machine, operators typically learn through on-the-job training. Indeed, learning by doing is in general more effective, tasks done practically are remembered more easily, and the training is more motivating and less tiresome. On the other hand, this training method has several negative factors. In particular, safety may be a major issue in some training situations. An approach that may contribute overcoming negative factors is using Virtual Reality and digital simulations techniques for operators training. The research work presented in this paper concerns the development of a multisensory Virtual Reality environment for training operators to properly use machinery and Personal Protective Equipment (PPE). The context selected for the study is a university laboratory hosting manufacturing machinery. It has been developed an application that allows user to navigate the laboratory, to approach a machine and learn about how to operate it and also what PPE to use while operating. Specifically, the paper describes the design and implementation of the application.
{"title":"Multisensory VR for Delivering Training Content to Machinery Operators","authors":"M. Bordegoni, M. Carulli, E. Spadoni","doi":"10.1115/detc2021-69974","DOIUrl":"https://doi.org/10.1115/detc2021-69974","url":null,"abstract":"\u0000 The issue of training operators in the use of machinery is topical in the industrial field and in many other contexts, such as university laboratories. Training is about learning how to use machinery properly and safely.\u0000 Beyond the possibility of studying manuals to learn how to use a machine, operators typically learn through on-the-job training. Indeed, learning by doing is in general more effective, tasks done practically are remembered more easily, and the training is more motivating and less tiresome. On the other hand, this training method has several negative factors. In particular, safety may be a major issue in some training situations.\u0000 An approach that may contribute overcoming negative factors is using Virtual Reality and digital simulations techniques for operators training.\u0000 The research work presented in this paper concerns the development of a multisensory Virtual Reality environment for training operators to properly use machinery and Personal Protective Equipment (PPE). The context selected for the study is a university laboratory hosting manufacturing machinery. It has been developed an application that allows user to navigate the laboratory, to approach a machine and learn about how to operate it and also what PPE to use while operating. Specifically, the paper describes the design and implementation of the application.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"34 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82511748","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}
A. Kulkarni, P. Bhatt, Alec Kanyuck, Satyandra K. Gupta
Robotic Wire Arc Additive Manufacturing (WAAM) is the layer-by-layer deposition of molten metal to build a three-dimensional part. In this process, the fed metal wire is melted using an electric arc as a heat source. The process is sensitive to the arc conditions, such as arc length. While building WAAM parts, the metal beads overlap at corners causing material accumulation. Material accumulation is undesirable as it leads to uneven build height and process failures caused by arc length variation. This paper introduces a deposition speed regulation scheme to avoid the corner accumulation problem and build parts with uniform build height. The regulated speed has a complex relationship with the corner angle, bead geometry, and molten metal dynamics. So we need to train a model that can predict suitable speed regulations for corner angles encountered while building the part. We develop an unsupervised learning technique to characterize the uniformity of the bead profile of a WAAM built layer and check for anomalous bead profiles. We train a model using these results that can predict suitable speed regulation parameters for different corner angles. We test this model by building a WAAM part using our speed regulation scheme and validate if the built part has uniform build height and reduced corner defects.
{"title":"Using Unsupervised Learning for Regulating Deposition Speed During Robotic Wire Arc Additive Manufacturing","authors":"A. Kulkarni, P. Bhatt, Alec Kanyuck, Satyandra K. Gupta","doi":"10.1115/detc2021-71865","DOIUrl":"https://doi.org/10.1115/detc2021-71865","url":null,"abstract":"\u0000 Robotic Wire Arc Additive Manufacturing (WAAM) is the layer-by-layer deposition of molten metal to build a three-dimensional part. In this process, the fed metal wire is melted using an electric arc as a heat source. The process is sensitive to the arc conditions, such as arc length. While building WAAM parts, the metal beads overlap at corners causing material accumulation. Material accumulation is undesirable as it leads to uneven build height and process failures caused by arc length variation. This paper introduces a deposition speed regulation scheme to avoid the corner accumulation problem and build parts with uniform build height. The regulated speed has a complex relationship with the corner angle, bead geometry, and molten metal dynamics. So we need to train a model that can predict suitable speed regulations for corner angles encountered while building the part. We develop an unsupervised learning technique to characterize the uniformity of the bead profile of a WAAM built layer and check for anomalous bead profiles. We train a model using these results that can predict suitable speed regulation parameters for different corner angles. We test this model by building a WAAM part using our speed regulation scheme and validate if the built part has uniform build height and reduced corner defects.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"57 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88286601","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}
Zhuo Yang, Yan Lu, Simin Li, Jennifer Li, Yande Ndiaye, Hui Yang, S. Krishnamurty
To accelerate the adoption of Metal Additive Manufacturing (MAM) for production, an understanding of MAM process-structure-property (PSP) relationships is indispensable for quality control. A multitude of physical phenomena involved in MAM necessitates the use of multi-modal and in-process sensing techniques to model, monitor and control the process. The data generated from these sensors and process actuators are fused in various ways to advance our understanding of the process and to estimate both process status and part-in-progress states. This paper presents a hierarchical in-process data fusion framework for MAM, consisting of pointwise, trackwise, layerwise and partwise data analytics. Data fusion can be performed at raw data, feature, decision or mixed levels. The multi-scale data fusion framework is illustrated in detail using a laser powder bed fusion process for anomaly detection, material defect isolation, and part quality prediction. The multi-scale data fusion can be generally applied and integrated with real-time MAM process control, near-real-time layerwise repairing and buildwise decision making. The framework can be utilized by the AM research and standards community to rapidly develop and deploy interoperable tools and standards to analyze, process and exploit two or more different types of AM data. Common engineering standards for AM data fusion systems will dramatically improve the ability to detect, identify and locate part flaws, and then derive optimal policies for process control.
{"title":"In-Process Data Fusion for Process Monitoring and Control of Metal Additive Manufacturing","authors":"Zhuo Yang, Yan Lu, Simin Li, Jennifer Li, Yande Ndiaye, Hui Yang, S. Krishnamurty","doi":"10.1115/detc2021-71813","DOIUrl":"https://doi.org/10.1115/detc2021-71813","url":null,"abstract":"\u0000 To accelerate the adoption of Metal Additive Manufacturing (MAM) for production, an understanding of MAM process-structure-property (PSP) relationships is indispensable for quality control. A multitude of physical phenomena involved in MAM necessitates the use of multi-modal and in-process sensing techniques to model, monitor and control the process. The data generated from these sensors and process actuators are fused in various ways to advance our understanding of the process and to estimate both process status and part-in-progress states. This paper presents a hierarchical in-process data fusion framework for MAM, consisting of pointwise, trackwise, layerwise and partwise data analytics. Data fusion can be performed at raw data, feature, decision or mixed levels. The multi-scale data fusion framework is illustrated in detail using a laser powder bed fusion process for anomaly detection, material defect isolation, and part quality prediction. The multi-scale data fusion can be generally applied and integrated with real-time MAM process control, near-real-time layerwise repairing and buildwise decision making. The framework can be utilized by the AM research and standards community to rapidly develop and deploy interoperable tools and standards to analyze, process and exploit two or more different types of AM data. Common engineering standards for AM data fusion systems will dramatically improve the ability to detect, identify and locate part flaws, and then derive optimal policies for process control.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"47 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85102017","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}
We introduce a method to help protect against and mitigate possible consequences of major regional and global events that can disrupt a system design and manufacturing process. The method is intended to be used during the conceptual phase of system design when functional models have been developed and component solutions are being chosen. Disruptive events such as plane crashes killing many engineers from one company traveling together, disease outbreaks killing or temporarily disabling many people associated with one industrial sector who travel to the same conference regularly, geopolitical events that impose tariffs or complete cessation of trade with a country that supplies a critical component, and many other similar physical and virtual events can significantly delay or disrupt a system design process. By comparing alternative embodiment, component, and low-level functional solutions, solutions can be identified that better pass the bus factor where no one disruptive event will cause a major delay or disruption to a system design and manufacturing process. We present a simplified case study of a renewable energy generation and storage system intended for residential use to demonstrate the method. While some challenges to immediate adoption by practitioners exist, we believe the method has the potential to significantly improve system design processes so that systems are designed, manufactured, and delivered on schedule and on budget from the perspective of significant disruptive events to design and manufacturing.
{"title":"The Bus Factor in Conceptual System Design: Protecting a Design Process Against Major Regional and World Events","authors":"Douglas L. Van Bossuyt, R. Arlitt","doi":"10.1115/detc2021-70476","DOIUrl":"https://doi.org/10.1115/detc2021-70476","url":null,"abstract":"\u0000 We introduce a method to help protect against and mitigate possible consequences of major regional and global events that can disrupt a system design and manufacturing process. The method is intended to be used during the conceptual phase of system design when functional models have been developed and component solutions are being chosen. Disruptive events such as plane crashes killing many engineers from one company traveling together, disease outbreaks killing or temporarily disabling many people associated with one industrial sector who travel to the same conference regularly, geopolitical events that impose tariffs or complete cessation of trade with a country that supplies a critical component, and many other similar physical and virtual events can significantly delay or disrupt a system design process. By comparing alternative embodiment, component, and low-level functional solutions, solutions can be identified that better pass the bus factor where no one disruptive event will cause a major delay or disruption to a system design and manufacturing process. We present a simplified case study of a renewable energy generation and storage system intended for residential use to demonstrate the method. While some challenges to immediate adoption by practitioners exist, we believe the method has the potential to significantly improve system design processes so that systems are designed, manufactured, and delivered on schedule and on budget from the perspective of significant disruptive events to design and manufacturing.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"174 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73164169","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}
Britta Hale, Douglas L. Van Bossuyt, N. Papakonstantinou, B. O’Halloran
Fuelled by recent technological advances, Machine Learning (ML) is being introduced to safety and security-critical applications like defence systems, financial systems, and autonomous machines. ML components can be used either for processing input data and/or for decision making. The response time and success rate demands are very high and this means that the deployed training algorithms often produce complex models that are not readable and verifiable by humans (like multi layer neural networks). Due to the complexity of these models, achieving complete testing coverage is in most cases not realistically possible. This raises security threats related to the ML components presenting unpredictable behavior due to malicious manipulation (backdoor attacks). This paper proposes a methodology based on established security principles like Zero-Trust and defence-in-depth to help prevent and mitigate the consequences of security threats including ones emerging from ML-based components. The methodology is demonstrated on a case study of an Unmanned Aerial Vehicle (UAV) with a sophisticated Intelligence, Surveillance, and Reconnaissance (ISR) module.
{"title":"A Zero-Trust Methodology for Security of Complex Systems With Machine Learning Components","authors":"Britta Hale, Douglas L. Van Bossuyt, N. Papakonstantinou, B. O’Halloran","doi":"10.1115/detc2021-70442","DOIUrl":"https://doi.org/10.1115/detc2021-70442","url":null,"abstract":"\u0000 Fuelled by recent technological advances, Machine Learning (ML) is being introduced to safety and security-critical applications like defence systems, financial systems, and autonomous machines. ML components can be used either for processing input data and/or for decision making. The response time and success rate demands are very high and this means that the deployed training algorithms often produce complex models that are not readable and verifiable by humans (like multi layer neural networks). Due to the complexity of these models, achieving complete testing coverage is in most cases not realistically possible. This raises security threats related to the ML components presenting unpredictable behavior due to malicious manipulation (backdoor attacks). This paper proposes a methodology based on established security principles like Zero-Trust and defence-in-depth to help prevent and mitigate the consequences of security threats including ones emerging from ML-based components. The methodology is demonstrated on a case study of an Unmanned Aerial Vehicle (UAV) with a sophisticated Intelligence, Surveillance, and Reconnaissance (ISR) module.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"16 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86547297","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}
The importance of process planning in manufacturing systems has been widely recognized. Process planning is a complex process with abundant information, intensive knowledge and miscellaneous experience, which leads to many challenges in its innovation. Innovation has been supported by innovative methods and information integration. However, the existing research on innovative design theory and information integration only focused on product design but rarely process planning. Therefore, it is urgent to study how to systematically use multi-dimensional information to influence process planning and realize innovative process design. An model for innovative process planning was proposed. And the general strategy of process innovation design was established. The process knowledge management and application methods were put forward, and the knowledge service system for innovative process was established. In order to support process innovation in practice, the framework of innovative process design service platform was established. Finally, a case is taken to illustrate the feasibility of the proposed method. The result shows that the proposed method can guide the designers to produce innovative process plans rapidly to solve the shortcomings of traditional process plans.
{"title":"Research on Multi-Dimensional Information Service Oriented to Innovative Process Planning","authors":"Jun Li, Xin Guo, Wu Zhao","doi":"10.1115/detc2021-71137","DOIUrl":"https://doi.org/10.1115/detc2021-71137","url":null,"abstract":"\u0000 The importance of process planning in manufacturing systems has been widely recognized. Process planning is a complex process with abundant information, intensive knowledge and miscellaneous experience, which leads to many challenges in its innovation. Innovation has been supported by innovative methods and information integration. However, the existing research on innovative design theory and information integration only focused on product design but rarely process planning. Therefore, it is urgent to study how to systematically use multi-dimensional information to influence process planning and realize innovative process design. An model for innovative process planning was proposed. And the general strategy of process innovation design was established. The process knowledge management and application methods were put forward, and the knowledge service system for innovative process was established. In order to support process innovation in practice, the framework of innovative process design service platform was established. Finally, a case is taken to illustrate the feasibility of the proposed method. The result shows that the proposed method can guide the designers to produce innovative process plans rapidly to solve the shortcomings of traditional process plans.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"38 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85603644","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}
The several loops characterizing the design process used to slow down the development of new projects. Since the 70s, the design process has changed due to the new technologies and tools related to Computer-Aided Design software and Virtual Reality applications that make almost the whole process digital. However, the concept phase of the design process is still based on traditional approaches, while digital tools are poor exploited. In this phase, designers need tools that allow them to rapidly save and freeze their ideas, such as sketching on paper, which is not integrated in the digital-based process. The paper presents a new gestural interface to give designers more support by introducing an effective device for 3D modelling to improve and speed up the conceptual design process. We designed a set of gestures to allow people from different background to 3D model their ideas in a natural way. A testing session with 17 participants allowed us to verify if the proposed interaction was intuitive or not. At the end of the tests, all participants succeeded in the 3D modelling of a simple shape (a column) by only using air gestures in a relatively short amount of time exactly how they expected it to be built, confirming the proposed interaction.
{"title":"Gestural Interfaces to Support the Sketching Activities of Designers","authors":"Pierstefano Bellani, M. Carulli, G. Caruso","doi":"10.1115/detc2021-71233","DOIUrl":"https://doi.org/10.1115/detc2021-71233","url":null,"abstract":"\u0000 The several loops characterizing the design process used to slow down the development of new projects. Since the 70s, the design process has changed due to the new technologies and tools related to Computer-Aided Design software and Virtual Reality applications that make almost the whole process digital. However, the concept phase of the design process is still based on traditional approaches, while digital tools are poor exploited. In this phase, designers need tools that allow them to rapidly save and freeze their ideas, such as sketching on paper, which is not integrated in the digital-based process.\u0000 The paper presents a new gestural interface to give designers more support by introducing an effective device for 3D modelling to improve and speed up the conceptual design process. We designed a set of gestures to allow people from different background to 3D model their ideas in a natural way. A testing session with 17 participants allowed us to verify if the proposed interaction was intuitive or not. At the end of the tests, all participants succeeded in the 3D modelling of a simple shape (a column) by only using air gestures in a relatively short amount of time exactly how they expected it to be built, confirming the proposed interaction.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"15 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81838141","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}
In the past few years, remote learning has been on a trend of steady growth and it is projected to remain on that course in the years to come. Additionally, the global COVID-19 pandemic forced a shift to remote learning which accelerated the existing trend to remote education. Unfortunately, learners find remote classes less engaging than traditional face-to-face classes. One technology that has shown great potential to improve students’ engagement, both in face-to-face classes and remote classes, is Virtual Reality (VR). Nevertheless, while educators are no longer limited to expensive, high-tech, and high-fidelity VR hardware thanks to the introduction of low-cost, low fidelity headset, like the Google Cardboard, educators are still limited in getting relevant content and find it difficult to create their own VR teaching modules. With the objective to address these limitations, this work introduces a new process to create VR content that is easy, rapid, and affordable for educators to adopt and implement into their curriculum. The results indicate great potential for low-cost VR in remote learning as the sample of students in this study reported that they enjoyed the ‘first-hand experience’ of touring places that were inaccessible to them due to the pandemic. However, the findings also show a strong need to address usability issues such as blurriness and dizziness.
{"title":"Creating Virtual Reality Teaching Modules for Low-Cost Headsets","authors":"Takudzwa Mujuru, C. López","doi":"10.1115/detc2021-72084","DOIUrl":"https://doi.org/10.1115/detc2021-72084","url":null,"abstract":"\u0000 In the past few years, remote learning has been on a trend of steady growth and it is projected to remain on that course in the years to come. Additionally, the global COVID-19 pandemic forced a shift to remote learning which accelerated the existing trend to remote education. Unfortunately, learners find remote classes less engaging than traditional face-to-face classes. One technology that has shown great potential to improve students’ engagement, both in face-to-face classes and remote classes, is Virtual Reality (VR). Nevertheless, while educators are no longer limited to expensive, high-tech, and high-fidelity VR hardware thanks to the introduction of low-cost, low fidelity headset, like the Google Cardboard, educators are still limited in getting relevant content and find it difficult to create their own VR teaching modules. With the objective to address these limitations, this work introduces a new process to create VR content that is easy, rapid, and affordable for educators to adopt and implement into their curriculum. The results indicate great potential for low-cost VR in remote learning as the sample of students in this study reported that they enjoyed the ‘first-hand experience’ of touring places that were inaccessible to them due to the pandemic. However, the findings also show a strong need to address usability issues such as blurriness and dizziness.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"43 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84147311","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}
Byeong-Min Roh, S. Kumara, Hui Yang, T. Simpson, P. Witherell, Yan Lu
Metal additive manufacturing (MAM) provides a larger design space with accompanying manufacturability than traditional manufacturing. Recently, much research has focused on simulating the MAM process with regards to part geometry, porosity, and microstructure properties. Despite continued advances, MAM processes have many variables that are not well understood with respect to their effect on the part quality. With the common use of in-situ sensors — such as CMOS cameras and infrared cameras — numerous, real-time datasets can be captured and analyzed for monitoring both the process and the part. However, currently, real-time data predominantly focuses on the build failure and process anomalies by capturing the printing defects (cracks/peel-off). A large amount of data — such as melt pool geometries and temperature gradients — are just beginning to be explored, along with their connections to final part quality. Towards investigating these connections, in this paper we propose models that capture numerous sensor capabilities and associate them with the corresponding, real-time, physical phenomena. These sensor models lay the foundation for a comprehensive, knowledge framework that forms the basis for quality monitoring and management of MAM process outcomes. Using our previously developed process ontology model [1–3], which describes the relationship between process variables and process outcomes, we can discover the relationship between the real-time, physical phenomena and the deviations in the targeted, build quality. For example, statistically significant sensor data that predicts deviations from targeted process qualities can be detected and used to control the process parameters. Case studies that scope the physical phenomena and sensor data are provided for verifying the effectiveness and efficiency of the proposed qualification and certification models.
{"title":"In-Situ Observation Selection for Quality Management in Metal Additive Manufacturing","authors":"Byeong-Min Roh, S. Kumara, Hui Yang, T. Simpson, P. Witherell, Yan Lu","doi":"10.1115/detc2021-70035","DOIUrl":"https://doi.org/10.1115/detc2021-70035","url":null,"abstract":"\u0000 Metal additive manufacturing (MAM) provides a larger design space with accompanying manufacturability than traditional manufacturing. Recently, much research has focused on simulating the MAM process with regards to part geometry, porosity, and microstructure properties. Despite continued advances, MAM processes have many variables that are not well understood with respect to their effect on the part quality. With the common use of in-situ sensors — such as CMOS cameras and infrared cameras — numerous, real-time datasets can be captured and analyzed for monitoring both the process and the part.\u0000 However, currently, real-time data predominantly focuses on the build failure and process anomalies by capturing the printing defects (cracks/peel-off). A large amount of data — such as melt pool geometries and temperature gradients — are just beginning to be explored, along with their connections to final part quality. Towards investigating these connections, in this paper we propose models that capture numerous sensor capabilities and associate them with the corresponding, real-time, physical phenomena. These sensor models lay the foundation for a comprehensive, knowledge framework that forms the basis for quality monitoring and management of MAM process outcomes.\u0000 Using our previously developed process ontology model [1–3], which describes the relationship between process variables and process outcomes, we can discover the relationship between the real-time, physical phenomena and the deviations in the targeted, build quality. For example, statistically significant sensor data that predicts deviations from targeted process qualities can be detected and used to control the process parameters. Case studies that scope the physical phenomena and sensor data are provided for verifying the effectiveness and efficiency of the proposed qualification and certification models.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"20 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78119038","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}
Jiaqi Lyu, Javid Akhavan Taheri Boroujeni, S. Manoochehri
Additive Manufacturing (AM) is a trending technology with great potential in manufacturing. In-situ process monitoring is a critical part of quality assurance for AM process. Anomalies need to be identified early to avoid further deterioration of the part quality. This paper presents an in-situ laser-based process monitoring and anomaly identification system to assure fabrication quality of Fused Filament Fabrication (FFF) machine. The proposed data processing and communication architecture of the monitoring system establishes the data transformation between workstation, FFF machine, and laser scanner control system. The data processing performs calibration, filtering, and segmentation for the point cloud of each layer acquired from a 3D laser scanner during the fabrication process. The point cloud dataset with in-plane surface depth information is converted into a 2D depth image. Each depth image is discretized into 100 equal regions of interest and then labeled accordingly. Using the image dataset, four Machine Learning (ML) classification models are trained and compared, namely Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Convolutional Neural Network (CNN), and Hybrid Convolution AutoEncoder (HCAE). The HCAE classification model shows the best performance based on F-scores to effectively classify the in-plane anomalies into four categories, namely empty region, normal region, bulge region, and dent region.
{"title":"In-Situ Laser-Based Process Monitoring and In-Plane Surface Anomaly Identification for Additive Manufacturing Using Point Cloud and Machine Learning","authors":"Jiaqi Lyu, Javid Akhavan Taheri Boroujeni, S. Manoochehri","doi":"10.1115/detc2021-69436","DOIUrl":"https://doi.org/10.1115/detc2021-69436","url":null,"abstract":"\u0000 Additive Manufacturing (AM) is a trending technology with great potential in manufacturing. In-situ process monitoring is a critical part of quality assurance for AM process. Anomalies need to be identified early to avoid further deterioration of the part quality. This paper presents an in-situ laser-based process monitoring and anomaly identification system to assure fabrication quality of Fused Filament Fabrication (FFF) machine. The proposed data processing and communication architecture of the monitoring system establishes the data transformation between workstation, FFF machine, and laser scanner control system. The data processing performs calibration, filtering, and segmentation for the point cloud of each layer acquired from a 3D laser scanner during the fabrication process. The point cloud dataset with in-plane surface depth information is converted into a 2D depth image. Each depth image is discretized into 100 equal regions of interest and then labeled accordingly. Using the image dataset, four Machine Learning (ML) classification models are trained and compared, namely Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Convolutional Neural Network (CNN), and Hybrid Convolution AutoEncoder (HCAE). The HCAE classification model shows the best performance based on F-scores to effectively classify the in-plane anomalies into four categories, namely empty region, normal region, bulge region, and dent region.","PeriodicalId":23602,"journal":{"name":"Volume 2: 41st Computers and Information in Engineering Conference (CIE)","volume":"34 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82027684","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：