2020 industryXchange Multi-Disciplinary Research Seed Grant Awardees

In collaboration with numerous research institutes and colleges, industryXchange has awarded eight Multi-Disciplinary Research Seed Grants to stimulate application-specific development in artificial intelligence (AI) and machine learning in the areas of cybersecurity, energy, healthcare, manufacturing, and transportation. Below is a list of the awardees along with an abstract of their planned research.

Industry partners interested in collaborating with any of the awarded faculty please contact Priya Baboo, pzb104@psu.edu.


Faculty: Youakim Badr, associate professor of data analytics; Partha Mukherjee, assistant professor of data analytics; Raghu Sangwan, associate professor of software engineering; and Satish Srinivasan, assistant professor of information sciences, all at Penn State Great Valley

Title: Managing Risks in AI Systems: Mitigating Vulnerabilities and Threats Using Design Tactics and Patterns

Abstract: Advances in artificial intelligence (AI) combined with sensors, actuators, and embedded systems have made it feasible to incorporate intelligence into software intensive-systems and safety-critical cyber physical systems. Not only should the correct functioning of these systems be tested at design and runtime, but so should their resilience to adversarial attacks and how taking precautionary steps to manage risks and vulnerabilities. Risks associated with poorly designed AI-systems that, if manifested, could expose them to potential threats. This project aims at addressing the challenge of mitigating vulnerabilities and threats in AI systems through a multidisciplinary approach, including expertise from risk management, monitoring distributed networks, software testing, fault detection, and adversarial countermeasures. By exploring the synergy between these disciplines, this project seeks to build a holistic framework to manage risks and demonstrates with a concrete scenario, how to manage potential threats in distributed AI video surveillance system.


Faculty: Rebecca Napolitano, assistant professor of architectural engineering, and Wesley Reinhart, assistant professor of materials science and engineering

Title: Enabling Energy-Efficient Building Envelopes: Artificial Intelligence for Nondestructive Evaluation

Abstract: Seventy-five percent of U.S. buildings will be retrofitted for energy efficiency by 2035. While it is important that the retrofitted buildings achieve necessary efficiency, this can be difficult since the state inside the building envelope is typically unknown. Radar methods have been developed to detect embedded elements (rebar and plumbing). However, these techniques can only provide qualitative information about inhomogeneities (voids, moisture). No techniques presently exist to extract interior images or models at a high resolution for use in simulation and monitoring. This work will use a hybrid physics-based simulation and artificial intelligence approach to infer the interior structure of an existing building envelope based on scattered radar signals. This novel, multidisciplinary approach will transform how energy-efficient retrofits and energy-usage monitoring are done. By providing a more accurate model of the envelope interior, it will enable predictive forecasting and optimized energy-retrofitting solutions leading to more energy-efficient infrastructure.


Faculty: Ankit Maheshwari, staff physician of PSHHVI electrophysiology at Penn State Health Milton S. Hershey Medical Center, and Vasant Honavar, professor and Edward Frymoyer Chair of Information Sciences and Technology

Title: Prediction of Subclinical Atrial Fibrillation in Patients with Cryptogenic Stroke using Point of Care Artificial Intelligence Guided Interpretation of Sinus Rhythm Electrocardiograms

Abstract: It is estimated that as many as 30% of cryptogenic strokes occur as a result of subclinical atrial fibrillation. Atrial fibrillation-related strokes can be significantly reduced by anticoagulant medications. Thus, patients suffering cryptogenic strokes are screened for atrial fibrillation with implantable heart rhythm monitors. Such monitors are expensive and result in significant healthcare costs. We aim to develop an artificial intelligence-based algorithm using convolution neural networks to predict the presence of atrial fibrillation from a sinus rhythm electrocardiogram, an inexpensive, ubiquitous clinical tool. If successful, patients suffering cryptogenic stroke will be able to receive effective point-of-care screening for atrial fibrillation to help inform treatment with anticoagulation earlier than the current strategy and in a more cost-effective way.


Faculty: Tarasankar Debroy, professor of materials science and engineering, and Todd Palmer, professor of engineering science and mechanics and materials science and engineering and director of Center for Innovative Sintered Products

Title: Machine learning based quality improvement of additively manufactured metallic components

Abstract: Metal printing is widely used in aerospace, automotive, and other industries because it can make complex parts that are difficult to produce by conventional manufacturing. However, the printed parts often suffer from defects and poor printability of feedstock materials. The current practice of solving these problems by prolonged trial-and-error selection of process variables results in high production cost and delay in part qualification. We seek to develop a machine learning-based solution for quality improvement of the printed metal parts by reducing defects. Replacement of the current empirical testing approach will reduce the time between the part design, production, and qualification in a cost-effective manner, minimize defects, and allow printing of new alloys. The proposed research at the cross-roads of metallurgy, mechanistic modeling, and machine learning will advance the evidence-based design of new printable alloys and enable efficient manufacturing process designs in and beyond additive manufacturing.


Faculty: Yiqi Zhang, assistant professor of industrial engineering, and Fenglong Ma, assistant professor of information sciences and technology

Title: Modeling of Driver Performance in Connected Vehicle Systems with Machine Learning Algorithms

Abstract: With the continued advances in connected vehicle technologies, the vehicles in the future connected vehicle system (CVS) will be equipped with the ability to communicate with each other. Such technology aims to provide drivers with safety-related information in a timely and reliable manner to improve transportation safety. Unfortunately, issues concerning the interaction between human drivers and connected vehicle systems in CVS design have not been fully recognized until recently. This project focuses on driver interaction behavior with CVS technologies and the development of machine learning models to predict driver performance in CVS with various traffic conditions. This work could be critical to quantify the impacts of CVS technologies on driver performance, optimize the design of in-vehicle safety applications with human-in-the-loop, and guide the development of warning algorithms for CVS technologies from a human factors perspective.


Faculty: Jinchao Xu, Verne M. Willaman Professor of Science, and John Yilin Wang, associate professor of petroleum and natural gas engineering

Title: Advanced and fast simulation technologies for modeling shale gas wells

Abstract: Expanded and successful development of natural gas resources in Pennsylvania has helped to secure energy for the nation and revive manufacturing industries including petrochemicals, chemicals, and steel. However, recovery ratio of shale gas in place is low, possibly less than 10%. To increase economic recovery of natural gas safely and in an environmentally friendly way, advanced tools are needed to analyze and evaluate the large amount of geological and engineering data associated with natural gas production. Current tools are limited in terms of physics, mechanisms, robustness, and computational efficiency. We propose to initiate a project on “Advanced and Fast Simulation Technologies for Modeling Shale Gas Wells” in collaborations with national labs and industrial partners to develop advanced numerical tools for other researchers and practicing engineers.


Faculty: Michael Lanagan, professor of engineering science and mechanics; Prasenjit Mitra, professor of information science and technology; and Ram Narayanan, professor of electrical engineering

Title: 5G Infrastructure and Materials Optimization Through Artificial Intelligence

Abstract: The 5G revolution will inspire entirely new applications in telecommunications, automotive, and health industries. Future building spaces and autonomous vehicles depend on high-speed communication for real-time connectivity and navigation, which will be achieved through high data rate and low latency 5G systems. In this seed project, artificial intelligence (AI) algorithms will be applied to office and home environments to learn about the impact of materials and structures on 5G communications. Large numbers of multiple-input-multiple-output (MIMO) cells will play an active role for increasing 5G capacity with additional enhancement of signal scattering through passive window and building structures. The technical objective of the project provides a new approach for understanding the roles of materials and structures on mm-wave scattering in complex and cluttered environments. The results of the work will be important for a diverse group of industries, including EM simulation companies, construction firms, and the materials suppliers.


Faculty: Penelope Kay Morrison, assistant professor of biobehavioral health at Penn State New Kensington; Kristin Sznajder, assistant professor of public health sciences at Penn State Health Milton S. Hershey Medical Center; Esther Obonyo, associate professor of engineering design and architectural engineering and director of the Global Building Network; and Soundar Kumara, Allen E. Pearce and Allen M. Pearce Professor of Industrial Engineering

Title: Healthy Residents 2.0: An exploratory pilot study of air quality and psychosocial well-being

Abstract: Air quality has been linked to poor physical health outcomes (e.g., asthma). The influence of air quality on psychosocial health outcomes (e.g. depression, isolation), however, remains unclear. This study aims to a) provide evidence for the effect of air pollutants on psychosocial health; b) develop a personalized adverse psychosocial health condition prediction model for early onset detection; c) improve our understanding of indoor/outdoor air condition interaction. Longitudinal self-assessment of psychosocial status and continuous sensory measurement of psychosocial markers will be paired with continuous sensory measurement of indoor and outdoor air quality. Combining these measurements can help predict deteriorating psychosocial health, identify specific air pollutants’ influence on psychosocial health, and track the development of those issues. Furthermore, personalized environmental and psychosocial monitoring datasets provide the opportunity leveraging machine learning techniques, together with participant assessments and observations, as a precision diagnostic approach to detecting more subtle psychosocial health conditions.