August 11, 2026 ME Faculty Mentor Students through Undergraduate Research Fellowships Written By: Kassi Akers Departments: Mechanical Engineering Categories: Awards | Research | Students | Undergraduate Department of Mechanical Engineering faculty/research staff will mentor 5 undergraduate students in independent research projects through university supported fellowships this academic year. With topics ranging from robotics to aerospace to engines, the collaborative efforts put forth through these projects will advance the educational experience and provide valuable growth for these undergraduate researchers. ’26-’27 Undergrad Research Fellowship Winners Sophomore Research Fellowships Tejas Dahiya Student: Tejas Dahiya Research Advisor: Saurabh Gupta Project Title: Predictive Sensor Health Index for Diesel Engine Emissions Monitoring This research develops a machine learning-based Sensor Health Index (SHI) to detect sensor degradation in diesel engines before failures result in emissions violations. Using neural networks trained on 217 operational samples from a 12L diesel testbed, the system predicts sensor outputs with a mean absolute percentage error of 4.3%, enabling early detection of eight failure modes, including drift, noise, bias, dropout, spikes, stuck-at faults, and saturation. The SHI quantifies sensor health on a scale from 0 to 1 and provides time-to-failure predictions. Multi-sensor monitoring distinguishes genuine sensor faults from normal engine transients, reducing false alarms. This framework addresses the environmental impact of undetected NOx sensor degradation, which can cause engines to over-inject fuel and increase CO₂ emissions by 5–10% per vehicle across commercial diesel fleets. Eve Lin Student: Eve Lin Research Advisors: Ramathasan Thevamaran Project Title: Additive Manufacturing Functionally Graded Foams to Control Local Mechanical Response Mechanical properties are essential to material design and selection in high-risk sectors, from biomedical to aerospace applications. Particularly in the space industry, structures such as satellites face issues including high-energy impacts from orbital strikes, thermal insulation, and vibration control. Currently, polymeric foams, a lightweight and relatively low-cost solution, are commonly used as thermal insulation and stiff sandwich wall cores to address these issues. These uniform-density foams are effective, yet have stochastic microstructure-dependent properties that are less controllable spatially. Functionally graded foams (FGFs) with controllable pore sizes present a better approach to fabricate materials with tunable mechanical and energy-related properties. Pore size control enables the creation of continuous density graded foams, which are difficult to produce using current production mechanisms but exhibit superior mechanical properties compared to discontinuous and uniform-density foams. In this study, functionally graded polyurethane foams will be created using direct ink writing (DIW), an additive manufacturing technique. DIW provides an efficient foam production technique that enables precise tuning of pore size and local behavior, and reduces inconsistencies that could impact mechanical performance. Samples of uniform density and varying density grades will undergo characterization and compressive testing to assess technique reproducibility and viability for tuning polymeric foam properties. Hilldale Undergraduate/Faculty Research Fellowships Alec Brey Student: Alec Brey Research Advisor: Harsh Sharma Project Title: Real-time control of soft robots via physics-preserving reduced-order models This project will develop a physics-preserving nonlinear model reduction framework for soft robotic systems to enable their real-time control. Soft robots exhibit highly nonlinear, high-dimensional dynamics that make model-based control computationally intractable. Drawing on ideas from Lagrangian mechanics, the project proposes a two-step learning strategy that first constructs physics-preserving linear reduced-order models via operator inference and then employs neural networks to learn nonlinear corrections while maintaining physical interpretability. The proposed approach will be tested on a swimming soft robot benchmark to evaluate accuracy and generalization. The resulting data-driven models will facilitate real-time control of soft robots and establish a general approach for highly flexible systems in structural engineering and aeroelasticity. Zoe Gureno Student: Zoe Gureno Research Advisor: Yunus Alapan Project: UV Surface Coatings for Magnetically Driven Soft Microrobotics Magnetically driven soft microrobotics’ are a unique developing technology with the ability to be coded with a magnetization profile that enables shape morphing and locomotion wirelessly controlled by external magnetic fields. Because of this maneuvering ability and remote actuation, the technology is gaining popularity in the healthcare field with the potential for precise drug delivery in the human body. As this technology becomes more sought after, researchers look to enhance both their mechanical performance and manufacturability to create more dependable and accessible forms of medicinal transfer. Mercapto group grafting via ultraviolet (UV) curing techniques provide potential routes of improvements for these characteristics. By grafting mercapto groups and long chain polymers to NdFeB magnetic microparticles, the produced chemical bonds and physical entanglement within the thiol-ene crosslinked PDMS system can increase toughness, printing resolution, and magnetic response. The proposed research aims to analyze the effects of UV-cured surface coatings on microrobot performance and their cross functionality with silica coated particles in order to maximize the efficiency and manufacturability of microrobots in the medical field. Student: Abigail Winn Research Advisor: Xiangru Xu Project: Scalable Verification and Experimental Validation of Neural Network-Enabled Thrust Vector Control for Reusable Rockets This project aims to develop a computationally efficient neural verification method for neural network-based control systems and validate predicted stability bounds on a physically unstable thrust-vector-control (TVC) testbed. Reusable launch vehicles and thrust-vector-controlled rockets are inherently unstable, where real-time control under actuator saturation, sensor noise, and modeling uncertainty is imperative to maintain safety. With the transition to reusable first-stage boosters that must survive multiple flights, clearly defined stability limits have become an increasingly important safety and economic priority. However, most robustness analyses for nonlinear control systems remain simulation-based, and neural network residual models further complicate verification due to the computational cost of scalable closed-loop analysis and the lack of formal validation on physical hardware. This work will develop a reachability algorithm that leverages structural properties to efficiently bound closed-loop behavior and compare predicted invariant sets against stability boundaries under actuator limits and delay, producing an experimentally evaluated verification framework for unstable aerospace systems.