School of Engineering Department of Mechanical and Aerospace Engineering 157 Department of Mechanical and Aerospace Engineering Design of Energy Conversion Device from Heat to Electricity Supervisor: CHEN Sherry / MAE Student: KWAN Chak Fung Kevin / MECH Course: UROP 1100, Spring When a train passes through a tunnel, the air flow induced by train would raise the temperature. With also the heat loss from the train, the temperature will decrease later. If such temperature vibration could be used to convert waste heat to electricity, it will reduce the energy loss from heat and improve the energy utilization. Therefore, this project aims to design a platform using pyroelectric materials for energy conversion from heat to electricity. Generative AI Pipeline for Material Design Supervisor: CHEN Sherry / MAE Student: LIE Ari Nathan Visesa / MECH Course: UROP 1100, Summer Ferroelectric materials are widely used in numerous applications, including energy storage, actuators and sensors. Development of this influential material is critical in the advancement of technology. However, these materials involve complex crystal structures requiring meticulous synthesis for studying and are accompanied by intrinsic properties which inhibit the materials effectiveness. As a result, the development of ferroelectric materials to overcome these limitations is hindered. Prediction of possible ferroelectric materials and their properties can greatly accelerate material development. Hence a database of ferroelectric materials to train the prediction algorithm is required. The project focuses on the development of a ferroelectric material database leveraging Large Language Models (LLMs) through generative AI pipelines. LLMs comprehension can facilitate the extraction of relevant ferroelectric material and their properties. The methodology of pipeline developments utilizes the LangChain framework. Evaluation of the pipelines indicate a requirement to subdivide the objective into discrete stages, facilitating key stages to be prompted. Additionally, incorporation of rationale and examples within the prompt optimizes response generated.
RkJQdWJsaXNoZXIy NDk5Njg=