Shuji Chang , Piyush Agarwal , Chris McCready , Luis Ricardez-Sandoval , Hector Budman
{"title":"基于灵敏度的自适应采样鲁棒PINN建模:最优传感器放置和结构不确定性处理的集成","authors":"Shuji Chang , Piyush Agarwal , Chris McCready , Luis Ricardez-Sandoval , Hector Budman","doi":"10.1016/j.jprocont.2025.103493","DOIUrl":null,"url":null,"abstract":"<div><div>Physics-informed neural networks (PINNs) have emerged as promising surrogate models for systems governed by partial differential equations (PDEs), yet their practical implementation faces challenges in training efficiency and robustness to uncertainties. This study refines and improves a recently proposed sensitivity-based adaptive sampling (SBS) methodology to address these challenges specifically for twice-differentiable PDE systems. We first conduct a systematic investigation of SBS hyper-parameters, including prediction horizon and adaptation rate, revealing their crucial role in training performance. To enhance PINN model robustness facing uncertainties, we propose two approaches: (1) incorporating sensor measurements at sensitivity-identified locations into the loss function, and (2) augmenting the PINN architecture with direct sensor data inputs. Results show that our proposed approaches achieve superior generalization capabilities and robustness. Furthermore, we demonstrate that the SBS methodology can serve for optimal sensor placement by identifying locations that maximize information gain for model training.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"152 ","pages":"Article 103493"},"PeriodicalIF":3.3000,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Robust PINN modeling via sensitivity-based adaptive sampling: Integration of optimal sensor placement and structural uncertainty handling\",\"authors\":\"Shuji Chang , Piyush Agarwal , Chris McCready , Luis Ricardez-Sandoval , Hector Budman\",\"doi\":\"10.1016/j.jprocont.2025.103493\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Physics-informed neural networks (PINNs) have emerged as promising surrogate models for systems governed by partial differential equations (PDEs), yet their practical implementation faces challenges in training efficiency and robustness to uncertainties. This study refines and improves a recently proposed sensitivity-based adaptive sampling (SBS) methodology to address these challenges specifically for twice-differentiable PDE systems. We first conduct a systematic investigation of SBS hyper-parameters, including prediction horizon and adaptation rate, revealing their crucial role in training performance. To enhance PINN model robustness facing uncertainties, we propose two approaches: (1) incorporating sensor measurements at sensitivity-identified locations into the loss function, and (2) augmenting the PINN architecture with direct sensor data inputs. Results show that our proposed approaches achieve superior generalization capabilities and robustness. Furthermore, we demonstrate that the SBS methodology can serve for optimal sensor placement by identifying locations that maximize information gain for model training.</div></div>\",\"PeriodicalId\":50079,\"journal\":{\"name\":\"Journal of Process Control\",\"volume\":\"152 \",\"pages\":\"Article 103493\"},\"PeriodicalIF\":3.3000,\"publicationDate\":\"2025-07-05\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Process Control\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0959152425001210\",\"RegionNum\":2,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"AUTOMATION & CONTROL SYSTEMS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Process Control","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0959152425001210","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"AUTOMATION & CONTROL SYSTEMS","Score":null,"Total":0}
Robust PINN modeling via sensitivity-based adaptive sampling: Integration of optimal sensor placement and structural uncertainty handling
Physics-informed neural networks (PINNs) have emerged as promising surrogate models for systems governed by partial differential equations (PDEs), yet their practical implementation faces challenges in training efficiency and robustness to uncertainties. This study refines and improves a recently proposed sensitivity-based adaptive sampling (SBS) methodology to address these challenges specifically for twice-differentiable PDE systems. We first conduct a systematic investigation of SBS hyper-parameters, including prediction horizon and adaptation rate, revealing their crucial role in training performance. To enhance PINN model robustness facing uncertainties, we propose two approaches: (1) incorporating sensor measurements at sensitivity-identified locations into the loss function, and (2) augmenting the PINN architecture with direct sensor data inputs. Results show that our proposed approaches achieve superior generalization capabilities and robustness. Furthermore, we demonstrate that the SBS methodology can serve for optimal sensor placement by identifying locations that maximize information gain for model training.
期刊介绍:
This international journal covers the application of control theory, operations research, computer science and engineering principles to the solution of process control problems. In addition to the traditional chemical processing and manufacturing applications, the scope of process control problems involves a wide range of applications that includes energy processes, nano-technology, systems biology, bio-medical engineering, pharmaceutical processing technology, energy storage and conversion, smart grid, and data analytics among others.
Papers on the theory in these areas will also be accepted provided the theoretical contribution is aimed at the application and the development of process control techniques.
Topics covered include:
• Control applications• Process monitoring• Plant-wide control• Process control systems• Control techniques and algorithms• Process modelling and simulation• Design methods
Advanced design methods exclude well established and widely studied traditional design techniques such as PID tuning and its many variants. Applications in fields such as control of automotive engines, machinery and robotics are not deemed suitable unless a clear motivation for the relevance to process control is provided.