Vlado Menkovski bio photo

Vlado Menkovski

Assistant Professor at Eindhoven University of Technology.

Email Google Scholar

TLDR; This project is about the production of projection light engines. The goal is to detect faulty products in real time, before they leave the factory, to prevent them from malfunctioning at the customer side. A dataset containing a lot of production information for the products (such as geometrical data and melting process data) is available and is already being used to address this problem from a data science perspective. The goal of this assignment will be to try a different approach instead, using (X-ray) images of the bulbs and relate these images to process variations, with the purpose of detecting faulty or suspicious products.

Project Description: The goal of this assignment is to investigate how (visual light and/or X-ray) images of projector light bulbs can aid in the detection of faulty or suspicious (prone to failure) light engines before they leave the factory, and preferably even in rejecting the half fabricated items in real-time.

It is suggested that the shape of the light engine might be an indicator of built-in stresses or micro-defects, and is hence a risk of later failure. The product shape is the result of a complex interaction of parts positioning and high temperature processes. Because of the high dimensionality of the images, the idea is to investigate the use of deep learning methods for visual data analysis.

An important end goal is root cause analysis; providing insights into why the light engines fail, possibly leading to suggestions for improvements in the production process and hence also leading to yield increase.

Besides the images, a dataset is available that contains information from the production process (such as machine/process settings). This data is already being used to investigate methods for fault detection based on data anomaly analysis. A combined approach with this data, as well as the images, may however lead to better results. The task is to investigate if and how both datasets can be combined to target this problem.

Signify will provide a dataset of images of the light engines from the manufacturing line, paired with data about the production process of these light bulbs (containing potentially up to 500+ features).

Company Description: Signify is the world leader in connected LED lighting products, systems and services. Through our innovations, we unlock the extraordinary potential of light for brighter lives and a better world. We serve professional and consumer markets, transforming urban spaces, communities, work places, stadiums, buildings, shopping centres and homes. Our products, systems and services help our customers to maximise energy use, drive efficiencies and deliver new experiences and services.