UROP Proceedings 2020-21

School of Engineering Department of Mechanical and Aerospace Engineering 172 Application of AI-based Technique to Enhance Thermal Comfort Sensing for Smart Air Conditioner Supervisor: LEE Yi-Kuen / MAE Student: CHOW Hong Kiu / IIM Course: UROP1000, Summer This project aims to develop a low-cost system detecting thermal comfort based on the predicted mean vote(PMV) and predicted percentage of dissatisfied (PPD) with application of Arduino, MIT App Inventor and sensors including temperature and humidity sensor, air velocity sensor and motion sensor. Due to limitation of current PMV model, which is low accuracy, an improvement on the current model is worth considering. A survey is conducted to compare the difference between PMV and PPD based on current equation and observed mean vote (OMV) and observed percentage of unacceptability (OPU) to access the accuracy of the model and factor affecting the accuracy. I am involved in the Arduino coding and connection to sensors, MIT App Inventor on IOS device and the conduction of survey. Application of AI-based Technique to Enhance Thermal Comfort Sensing for Smart Air Conditioner Supervisor: LEE Yi-Kuen / MAE Student: LIM Her Wei / SENG Course: UROP1000, Summer Heating, Ventilating, and Air-Conditioning Systems (HVAC) use 24% of energy consumption in the commercial sector, according to an EMSD research article (CLP Power, 2021). However, these systems are not very efficient, causing a waste in both energy and lower thermal comfort, which is the main goal to improve in this research. To better the method of controlling HVACs will be crucial for reducing energy consumption and improving the comfort level of users. With this said, a smartphone app can be developed for the calculation and surveying of PMV, facilitating the process and increasing efficiency, while making a foundation for further smart-HVAC projects. Application of AI-based Technique to Enhance Thermal Comfort Sensing for Smart Air Conditioner Supervisor: LEE Yi-Kuen / MAE Student: SALIM Richie Wee / MATH-CS Course: UROP1000, Summer This paper will report the result of a data model created from machine learning methods using human pose data that has been collected before. Machine learning is used to compare the conventional models of equations and get the accuracy of the new model given the data. This is a crucial step in determining the user’s metabolism rate based on their current activity, so we can later determine the PMV for further processing. The model inputs our three-dimensional position processes it and classifies it into one of the basic human activities e.g. sitting, walking, running, or standing. Based on a widely used machine learning platform, the model is optimized for classification and reduces losses during training, resulting in a good performance and data accuracy.

RkJQdWJsaXNoZXIy NDk5Njg=