Optimizing Production Through Early Fault Detection in Smart Connected Factories

Smart connected factories are utilizing big data collected from IoT (Internet of Things) sensors to bring about a new industrial revolution. By developing classification models that utilize this data to predict product defects on the assembly line, businesses can reduce expenses and optimize performance. Identifying these parts early in the manufacturing process will result less wasting of raw materials or production time on products that will not pass final quality tests. My capstone project focused on developing an approach to predicting failure early along the assembly line. Public data collected at German manufacturing company Bosch’s smart factory was utilized to develop predictive analytics models in order to validate the proof of concept and aid in the discovery of recommendations for an early fault detection system. The goal of the project was to provide best practices and implementation guidance, based on the act of creating models with the Bosch data and drawing from industry experience. This included considerations for handling datasets that are highly imbalanced between the number of passing and failing observations, techniques that proved useful in model development, and best methods to validate model performance.

“I was able to pursue a capstone topic that was relevant to my industry, which enabled me to work on a real world issue in my field. I utilized the skills I learned in many of my courses to both develop the code, as well as communicate my results effectively.”