Published: November 23, 2020
favorite0 forum0 poll54
By: Zachary Harris, California State Polytechnic University
Machine learning and Internet of Things applications offer intelligent and efficient solutions in a variety of scenarios. Foresight and predictive capabilities are highly sought characteristics in emerging technology. However, these capabilities are most often achieved using purely software-based methods. Thus, these methods are at a disadvantage since they have no physical interaction with the technology and goods they analyze. This project investigates the utilization of a physical device for machine learning and internet of things applications. This project will examine the enhanced performance and efficiency of a system that creates predictions using information gathered from a physical device. This project is conducted using an intelligent tray focused on obtaining data and measurements using a customized force sensitive resistor. A Raspberry Pi is utilized to gather measurements and is utilized as a server and data base.
A functioning and usable model for the tray was produced which included a customized FSR that met the sensitivity requirements of the project. The customized FSR was able to measure the difference of a single coffee bean, however it also proved too volatile for long-term use and ultimately a manufactured, off-the-self FSR was installed. The simple linear regression model that was implemented was able to accurately predict the completion of the coffee in the jar and create an alert a user-specified number of days before completion. The web interface which was designed for the project worked very well and was able successfully to relay, download, and display the measurements, statuses, and predictions relating to the data and linear regression model. The web interface allowed this information to be readily available for any user to view and utilize at any time.