Online detection of wave soldering is an important method of inspecting defective products in the workshop. Accurate quality detection can reduce production costs and provide support for quality warning in wave soldering process. However, there are still problems of improving the detection accuracy for defect class. Although class imbalance in data can be addressed by data level methods such as over-sampling and under-sampling, these methods destroy the integrity of the original data set and may cause information loss and overfitting problems. In order to solve the above problems, this article focuses on how to design a new loss function that fuses class weights from focal loss (FS) and sample weights form AdaBoost to improve attention to the minority samples without changing data distribution. In this way, a FS-AdaBoost-RegNet model based on transfer learning is constructed to enhance the detection accuracy in industrial environment. Finally, the images of the wave soldering from an electronic assembly workshop are taken to validate the performance of the proposed method. The experiment on 941 testing samples of the imbalance datasets showed that the FS-AdaBoost-RegNet model with new loss function reached the overall accuracy of 98.39%, the overall recall of 96.19%. The results proved that the proposed method promotes the ability to identify defect class compared with other methods
{"title":"An online quality detection method with ensemble learning on imbalance data for wave soldering","authors":"Hanpeng Gao, Yu Guo, Shaohua Huang, Jian Xie, Daoyuan Liu, Tao Wu, Xu Tian","doi":"10.1115/1.4063068","DOIUrl":"https://doi.org/10.1115/1.4063068","url":null,"abstract":"\u0000 Online detection of wave soldering is an important method of inspecting defective products in the workshop. Accurate quality detection can reduce production costs and provide support for quality warning in wave soldering process. However, there are still problems of improving the detection accuracy for defect class. Although class imbalance in data can be addressed by data level methods such as over-sampling and under-sampling, these methods destroy the integrity of the original data set and may cause information loss and overfitting problems. In order to solve the above problems, this article focuses on how to design a new loss function that fuses class weights from focal loss (FS) and sample weights form AdaBoost to improve attention to the minority samples without changing data distribution. In this way, a FS-AdaBoost-RegNet model based on transfer learning is constructed to enhance the detection accuracy in industrial environment. Finally, the images of the wave soldering from an electronic assembly workshop are taken to validate the performance of the proposed method. The experiment on 941 testing samples of the imbalance datasets showed that the FS-AdaBoost-RegNet model with new loss function reached the overall accuracy of 98.39%, the overall recall of 96.19%. The results proved that the proposed method promotes the ability to identify defect class compared with other methods","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45375371","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}
Various sources of error can lead to the position accuracy of the robot being orders of magnitude worse than its repeatability. For the accuracy of drilling in the aviation field, high-precision assembly, and other fields depend on the industrial robot's absolute positioning accuracy, it is essential to improve the accuracy of absolute positioning by calibration. In the present paper, an error model of the robot is established considering both constant and joint-dependent kinematic errors, and the robot model is modified by the Hermite polynomial. To identify joint-dependent kinematic errors, a robot calibration method based on back-propagation neural network(BP) optimized by Sparrow Search Algorithm (SSA-BP) is proposed, which optimize the uncertainty of weights and thresholds in the BP algorithm . To validate the efficiency of the proposed method, experiments on an EFORT ECR5 robot were implemented. The positioning error is reduced from 3.1704 mm to 0.2798 mm, and the positioning accuracy is improved by 91.27%. With the new calibration method using SSA-BP, robot positioning errors can be effectively compensated for and the robot positioning accuracy can be improved significantly.
{"title":"Enhancing Robot Calibration through Reliable High-Order Hermite Polynomials Model and SSA-BP Optimization","authors":"Yujie Zhang, Qi Fang, Yu Xie, Weijie Zhang, Runxiang Yu","doi":"10.1115/1.4063035","DOIUrl":"https://doi.org/10.1115/1.4063035","url":null,"abstract":"\u0000 Various sources of error can lead to the position accuracy of the robot being orders of magnitude worse than its repeatability. For the accuracy of drilling in the aviation field, high-precision assembly, and other fields depend on the industrial robot's absolute positioning accuracy, it is essential to improve the accuracy of absolute positioning by calibration. In the present paper, an error model of the robot is established considering both constant and joint-dependent kinematic errors, and the robot model is modified by the Hermite polynomial. To identify joint-dependent kinematic errors, a robot calibration method based on back-propagation neural network(BP) optimized by Sparrow Search Algorithm (SSA-BP) is proposed, which optimize the uncertainty of weights and thresholds in the BP algorithm . To validate the efficiency of the proposed method, experiments on an EFORT ECR5 robot were implemented. The positioning error is reduced from 3.1704 mm to 0.2798 mm, and the positioning accuracy is improved by 91.27%. With the new calibration method using SSA-BP, robot positioning errors can be effectively compensated for and the robot positioning accuracy can be improved significantly.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44355432","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}
The connectivity of modern vehicles allows for the monitoring and analysis of a large amount of sensor data from vehicles during their normal operations. In recent years, there has been a growing interest in utilizing this data for the purposes of predictive maintenance. In this paper, a multi-label transfer learning approach is proposed using fourteen different pretrained classifier models retrained with engine simulation data to predict the failure conditions of a selected set of engine components. The retrained classifiers are designed such that the failure modes, including multimode failure, of an EGR, Compressor, Intercooler, and Fuel Injectors of a four-cylinder diesel engine can be identified. Time-series simulation data of various failure conditions, which includes performance degradation, is generated to retrain the classifier models to predict which components are failing at any given time. The test results of the retrained classifier models show that the overall classification performance is good, with the value of mean average precision varying from 0.7 to 0.75 for most retrained networks. To the best of the authors' knowledge, this work represents the first attempt to characterize such time-series data utilizing a multi-label deep learning approach.
{"title":"Determination of Multi-Component Failure in Automotive System using Deep Learning","authors":"John O'Donnell, Hwan-Sik Yoon","doi":"10.1115/1.4063003","DOIUrl":"https://doi.org/10.1115/1.4063003","url":null,"abstract":"\u0000 The connectivity of modern vehicles allows for the monitoring and analysis of a large amount of sensor data from vehicles during their normal operations. In recent years, there has been a growing interest in utilizing this data for the purposes of predictive maintenance. In this paper, a multi-label transfer learning approach is proposed using fourteen different pretrained classifier models retrained with engine simulation data to predict the failure conditions of a selected set of engine components. The retrained classifiers are designed such that the failure modes, including multimode failure, of an EGR, Compressor, Intercooler, and Fuel Injectors of a four-cylinder diesel engine can be identified. Time-series simulation data of various failure conditions, which includes performance degradation, is generated to retrain the classifier models to predict which components are failing at any given time. The test results of the retrained classifier models show that the overall classification performance is good, with the value of mean average precision varying from 0.7 to 0.75 for most retrained networks. To the best of the authors' knowledge, this work represents the first attempt to characterize such time-series data utilizing a multi-label deep learning approach.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"43558183","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}
Quantum computing as the emerging paradigm for scientific computing has attracted significant research attention in the past decade. Quantum algorithms to solve the problems of linear systems, eigenvalue, optimization, machine learning, and others have been developed. The main advantage of utilizing quantum computer to solve optimization problems is that quantum superposition allows for massive parallel searching of solutions. This article provides an overview of fundamental quantum algorithms that can be used to solve optimization problems, including Grover search, quantum phase estimation, quantum annealing, quantum approximate optimization algorithm, variational quantum eigensolver, and quantum walk. A review of recent applications of quantum optimization methods for engineering design, including materials design and topology optimization, is also given. The challenges to develop scalable and reliable quantum algorithms for engineering optimization are discussed.
{"title":"Opportunities and Challenges of Quantum Computing for Engineering Optimization","authors":"Yan Wang, Jungin E. Kim, K. Suresh","doi":"10.1115/1.4062969","DOIUrl":"https://doi.org/10.1115/1.4062969","url":null,"abstract":"\u0000 Quantum computing as the emerging paradigm for scientific computing has attracted significant research attention in the past decade. Quantum algorithms to solve the problems of linear systems, eigenvalue, optimization, machine learning, and others have been developed. The main advantage of utilizing quantum computer to solve optimization problems is that quantum superposition allows for massive parallel searching of solutions. This article provides an overview of fundamental quantum algorithms that can be used to solve optimization problems, including Grover search, quantum phase estimation, quantum annealing, quantum approximate optimization algorithm, variational quantum eigensolver, and quantum walk. A review of recent applications of quantum optimization methods for engineering design, including materials design and topology optimization, is also given. The challenges to develop scalable and reliable quantum algorithms for engineering optimization are discussed.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46728345","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}
Ananya Ipsita, Runlin Duan, Hao Li, Subramanian C, Yuanzhi Cao, Min Liu, Alexander J. Quinn, Karthik Ramani
Domain users (DUs) with a knowledge base in specialized fields are frequently excluded from authoring Virtual Reality (VR)-based applications in corresponding fields. This is largely due to the requirement of VR programming expertise needed to author these applications. To address this concern, we developed VRFromX, a system workflow design to make the virtual content creation process accessible to DUs irrespective of their programming skills and experience. VRFromX provides an in-situ process of content creation in VR that (a) allows users to select regions of interest in scanned point clouds or sketch in mid-air using a brush tool to retrieve virtual models, and (b) then attach behavioral properties to those objects. Using a welding use case, we performed a usability evaluation of VRFromX with 20 DUs from which 12 were novices in VR programming. Study results indicated positive user ratings for the system features with no significant differences across users with or without VR programming expertise. Based on the qualitative feedback, we also implemented two other use cases to demonstrate potential applications. We envision that the solution can facilitate the adoption of the immersive technology to create meaningful virtual environments.
{"title":"The Design of a Virtual Prototyping System for Authoring Interactive VR Environments from Real World Scans","authors":"Ananya Ipsita, Runlin Duan, Hao Li, Subramanian C, Yuanzhi Cao, Min Liu, Alexander J. Quinn, Karthik Ramani","doi":"10.1115/1.4062970","DOIUrl":"https://doi.org/10.1115/1.4062970","url":null,"abstract":"\u0000 Domain users (DUs) with a knowledge base in specialized fields are frequently excluded from authoring Virtual Reality (VR)-based applications in corresponding fields. This is largely due to the requirement of VR programming expertise needed to author these applications. To address this concern, we developed VRFromX, a system workflow design to make the virtual content creation process accessible to DUs irrespective of their programming skills and experience. VRFromX provides an in-situ process of content creation in VR that (a) allows users to select regions of interest in scanned point clouds or sketch in mid-air using a brush tool to retrieve virtual models, and (b) then attach behavioral properties to those objects. Using a welding use case, we performed a usability evaluation of VRFromX with 20 DUs from which 12 were novices in VR programming. Study results indicated positive user ratings for the system features with no significant differences across users with or without VR programming expertise. Based on the qualitative feedback, we also implemented two other use cases to demonstrate potential applications. We envision that the solution can facilitate the adoption of the immersive technology to create meaningful virtual environments.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47636829","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}
There is a growing interest in using deep learning technologies within the manufacturing industry to improve quality, productivity, safety, and efficiency, while also reducing costs and cycle time. This paper discusses the primary applications of deep learning currently being employed, including identifying defects during high-mix production, optimizing processes, streamlining the supply chain, predicting maintenance needs, and recognizing human activity. The paper offers a brief summary of the various components of deep learning technology and their roles. Additionally, the paper draws attention to the current challenges and limitations that need to be addressed to fully realize the potential of deep learning technology in manufacturing. Lastly, several future directions for research within the field are proposed to further improve the use of deep learning in manufacturing.
{"title":"The Role of Deep Learning in Manufacturing Applications: Challenges and Opportunities","authors":"R. Malhan, S. Gupta","doi":"10.1115/1.4062939","DOIUrl":"https://doi.org/10.1115/1.4062939","url":null,"abstract":"\u0000 There is a growing interest in using deep learning technologies within the manufacturing industry to improve quality, productivity, safety, and efficiency, while also reducing costs and cycle time. This paper discusses the primary applications of deep learning currently being employed, including identifying defects during high-mix production, optimizing processes, streamlining the supply chain, predicting maintenance needs, and recognizing human activity. The paper offers a brief summary of the various components of deep learning technology and their roles. Additionally, the paper draws attention to the current challenges and limitations that need to be addressed to fully realize the potential of deep learning technology in manufacturing. Lastly, several future directions for research within the field are proposed to further improve the use of deep learning in manufacturing.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75367032","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}
J. Allen, Anand Balu Nellippallil, Zhenjun Ming, J. Milisavljevic-Syed, F. Mistree
In the context of the theme for this special issue, namely, challenges and opportunities in computing research to enable next generation engineering applications, our intent in writing this paper is to seed the dialog on furthering computing research associated with the design of cyber-physical-social systems. Cyber-Physical-Social Systems (CPSS's) are natural extensions of Cyber-Physical Systems (CPS's) that add the consideration of human interactions and cooperation with cyber systems and physical systems. CPSS's are becoming increasingly important as we face challenges such as regulating our impact on the environment, eradicating disease, transitioning to digital and sustainable manufacturing, and improving healthcare. Human stakeholders in these systems are integral to the effectiveness of these systems. One of the key features of CPSS is that the form, structure, and interactions constantly evolve to meet changes in the environment. Design of evolving CPSS include making tradeoffs amongst the cyber, the physical, and the social systems. Advances in computing and information science have given us opportunities to ask difficult, and important questions, especially those related to cyber-physical-social systems. In this paper we identify research opportunities worth investigating. We start with theoretical and mathematical frameworks for identifying and framing the problem – specifically, problem identification and formulation, data management, CPSS modeling and CPSS in action. Then we discuss issues related to the design of CPSS including decision making, computational platform support, and verification and validation. Building on this foundation, we suggest a way forward.
{"title":"Designing Evolving Cyber-Physical-Social Systems: Computational Research Opportunities","authors":"J. Allen, Anand Balu Nellippallil, Zhenjun Ming, J. Milisavljevic-Syed, F. Mistree","doi":"10.1115/1.4062883","DOIUrl":"https://doi.org/10.1115/1.4062883","url":null,"abstract":"\u0000 In the context of the theme for this special issue, namely, challenges and opportunities in computing research to enable next generation engineering applications, our intent in writing this paper is to seed the dialog on furthering computing research associated with the design of cyber-physical-social systems. Cyber-Physical-Social Systems (CPSS's) are natural extensions of Cyber-Physical Systems (CPS's) that add the consideration of human interactions and cooperation with cyber systems and physical systems. CPSS's are becoming increasingly important as we face challenges such as regulating our impact on the environment, eradicating disease, transitioning to digital and sustainable manufacturing, and improving healthcare. Human stakeholders in these systems are integral to the effectiveness of these systems. One of the key features of CPSS is that the form, structure, and interactions constantly evolve to meet changes in the environment. Design of evolving CPSS include making tradeoffs amongst the cyber, the physical, and the social systems. Advances in computing and information science have given us opportunities to ask difficult, and important questions, especially those related to cyber-physical-social systems. In this paper we identify research opportunities worth investigating. We start with theoretical and mathematical frameworks for identifying and framing the problem – specifically, problem identification and formulation, data management, CPSS modeling and CPSS in action. Then we discuss issues related to the design of CPSS including decision making, computational platform support, and verification and validation. Building on this foundation, we suggest a way forward.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84937803","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}
Marek S. Lukasiewicz, M. Rossoni, E. Spadoni, Nicolò Dozio, M. Carulli, F. Ferrise, M. Bordegoni
As the Metaverse gains popularity due to its use in various industries, so does the desire to take advantage of all its potential. While visual and audio technologies already provide access to the Metaverse, there is increasing interest in haptic and olfactory technologies, which are less developed and have been studied for a shorter time. Currently, there are limited options for users to experience the olfactory aspect of the Metaverse. This paper introduces an open-source kit that makes it simple to add the sense of smell to the Metaverse. The details of the solution, including its technical specifications, are outlined to enable potential users to utilize, test, and enhance the project and make it available to the scientific community.
{"title":"An open-source Olfactory Display to add the sense of smell to the Metaverse","authors":"Marek S. Lukasiewicz, M. Rossoni, E. Spadoni, Nicolò Dozio, M. Carulli, F. Ferrise, M. Bordegoni","doi":"10.1115/1.4062889","DOIUrl":"https://doi.org/10.1115/1.4062889","url":null,"abstract":"\u0000 As the Metaverse gains popularity due to its use in various industries, so does the desire to take advantage of all its potential. While visual and audio technologies already provide access to the Metaverse, there is increasing interest in haptic and olfactory technologies, which are less developed and have been studied for a shorter time. Currently, there are limited options for users to experience the olfactory aspect of the Metaverse. This paper introduces an open-source kit that makes it simple to add the sense of smell to the Metaverse. The details of the solution, including its technical specifications, are outlined to enable potential users to utilize, test, and enhance the project and make it available to the scientific community.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45624163","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}
Jiangce Chen, Justin Pierce, Glen Williams, Timothy W. Simpson, N. Meisel, Sneha Prabha Narra, Christopher McComb
The temperature history of an additively-manufactured part plays a critical role in determining process-structure-property relationships in fusion-based additive manufacturing (AM) processes. Therefore, fast thermal simulation methods are needed for a variety of AM tasks, from temperature history prediction for part design and process planning to in-situ temperature monitoring and control during manufacturing. However, conventional numerical simulation methods fall short in satisfying the strict requirements of these applications due to the large space and time scales involved. While data-driven surrogate models are of interest for their rapid computation capabilities, the performance of these models relies on the size and quality of the training data, which is often prohibitively expensive to create. Physics-informed neural networks (PINNs) mitigate the need for large datasets by imposing physical principles during the training process. This work investigates the use of a PINN to predict the time-varying temperature distribution in a part during manufacturing with Laser Powder Bed Fusion (L-PBF). Notably, the use of the PINN in this study enables the model to be trained solely on randomly-synthesized data. This training data is both inexpensive to obtain and the presence of stochasticity in the dataset improves the generalizability of the trained model. Results show that the PINN model achieves higher accuracy than a comparable artificial neural network trained on labeled data. Further, the PINN model trained in this work maintains high accuracy in predicting temperature for laser path scanning strategies unseen in the training data.
{"title":"ACCELERATING THERMAL SIMULATIONS IN ADDITIVE MANUFACTURING BY TRAINING PHYSICS-INFORMED NEURAL NETWORKS WITH RANDOMLY-SYNTHESIZED DATA","authors":"Jiangce Chen, Justin Pierce, Glen Williams, Timothy W. Simpson, N. Meisel, Sneha Prabha Narra, Christopher McComb","doi":"10.1115/1.4062852","DOIUrl":"https://doi.org/10.1115/1.4062852","url":null,"abstract":"\u0000 The temperature history of an additively-manufactured part plays a critical role in determining process-structure-property relationships in fusion-based additive manufacturing (AM) processes. Therefore, fast thermal simulation methods are needed for a variety of AM tasks, from temperature history prediction for part design and process planning to in-situ temperature monitoring and control during manufacturing. However, conventional numerical simulation methods fall short in satisfying the strict requirements of these applications due to the large space and time scales involved. While data-driven surrogate models are of interest for their rapid computation capabilities, the performance of these models relies on the size and quality of the training data, which is often prohibitively expensive to create. Physics-informed neural networks (PINNs) mitigate the need for large datasets by imposing physical principles during the training process. This work investigates the use of a PINN to predict the time-varying temperature distribution in a part during manufacturing with Laser Powder Bed Fusion (L-PBF). Notably, the use of the PINN in this study enables the model to be trained solely on randomly-synthesized data. This training data is both inexpensive to obtain and the presence of stochasticity in the dataset improves the generalizability of the trained model. Results show that the PINN model achieves higher accuracy than a comparable artificial neural network trained on labeled data. Further, the PINN model trained in this work maintains high accuracy in predicting temperature for laser path scanning strategies unseen in the training data.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":3.1,"publicationDate":"2023-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44516192","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}
Abstract Semantic knowledge of part-part and part-whole relationships in assemblies is useful for a variety of tasks from searching design repositories to the construction of engineering knowledge bases. In this work, we propose that the natural language names designers use in computer aided design (CAD) software are a valuable source of such knowledge, and that large language models (LLMs) contain useful domain-specific information for working with this data as well as other CAD and engineering-related tasks. In particular, we extract and clean a large corpus of natural language part, feature, and document names and use this to quantitatively demonstrate that a pre-trained language model can outperform numerous benchmarks on three self-supervised tasks, without ever having seen this data before. Moreover, we show that fine-tuning on the text data corpus further boosts the performance on all tasks, thus demonstrating the value of the text data which until now has been largely ignored. We also identify key limitations to using LLMs with text data alone, and our findings provide a strong motivation for further work into multi-modal text-geometry models. To aid and encourage further work in this area we make all our data and code publicly available.
{"title":"What’s in a Name? Evaluating Assembly-Part Semantic Knowledge in Language Models Through User-Provided Names in Computer Aided Design Files","authors":"Peter Meltzer, Joseph Lambourne, Daniele Grandi","doi":"10.1115/1.4062454","DOIUrl":"https://doi.org/10.1115/1.4062454","url":null,"abstract":"Abstract Semantic knowledge of part-part and part-whole relationships in assemblies is useful for a variety of tasks from searching design repositories to the construction of engineering knowledge bases. In this work, we propose that the natural language names designers use in computer aided design (CAD) software are a valuable source of such knowledge, and that large language models (LLMs) contain useful domain-specific information for working with this data as well as other CAD and engineering-related tasks. In particular, we extract and clean a large corpus of natural language part, feature, and document names and use this to quantitatively demonstrate that a pre-trained language model can outperform numerous benchmarks on three self-supervised tasks, without ever having seen this data before. Moreover, we show that fine-tuning on the text data corpus further boosts the performance on all tasks, thus demonstrating the value of the text data which until now has been largely ignored. We also identify key limitations to using LLMs with text data alone, and our findings provide a strong motivation for further work into multi-modal text-geometry models. To aid and encourage further work in this area we make all our data and code publicly available.","PeriodicalId":54856,"journal":{"name":"Journal of Computing and Information Science in Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135904535","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：