Activiteiten

The case will guide you through the steps of an ASML metrology engineer trying to diagnose the issue with the new wafer heating correction model. You will be learning about the model’s purpose and inner workings, studying machine measurement data, and proposing experiments in an attempt to pinpoint the problem and fix it. About the case After two years of intensive study and development, you proudly present your new and improved software package, meant to suppress wafer heating effects in ASML’s new 140 million euro scanner. Just in time, too, since the scanners are about to be shipped to customers. And then, after one week of use, it turns out not to work. This happened to us a time ago, and led to a frantic and broad search for the root cause of the problem. During the investigation, we had to consider not just the software, but also the software’s relation to the surrounding mechatronic system, and the basic physics involved. The wafer heating student case is based on our records of this period. The case will guide you (in teams) through the steps of an ASML metrology engineer trying to diagnose the issue with the new wafer heating correction model. You will be learning about the model’s purpose and inner workings, studying machine measurement data, and proposing experiments in an attempt to pinpoint the problem and fix it. But don’t forget! The customer is already using the machine. You’d better hurry, because as always, time is money! Using a ‘choose your own adventure’ style game, diagnose a design problem in the interaction of the wafer heating feed-forward functionality. The game is played in teams which compete with each other to see who solves the problem while spending the lowest amount of ‘effort’ points. Schedule: 14:00 Introduction 14:30 Case 16:30 Wrap-up 17:00 Pizza at GEWIS This case is suitable for both mathematics and computer science students!

What do preliminary schools and garbage collection have in common? Indeed, both are areas where society benefits from mathematics! These days, lots of primary schools are digitizing and many are letting their pupils use tablets to solve exercises. Instead of letting them all make the same exercises, what if we can match exercises to a pupil's level? How do we determine then which exercises are difficult? And which pupils need more personal care? Mathematics will tell you how to solve this chicken and egg problem. Additionally, Marloes will show you how mathematics can save a municipality millions each year. But how do you convince a government that you are right and they are wrong (without literally saying this, believe me... that doesn't help). How to convince them to invest in your ideas and put trust in math. Join the lunch lecture and find out more about with garbage collection and maths! About Marloes Boswerger After her internship at LIME, Marloes finished her MSc (Industrial and Applied Mathematics) at the TU/e and joined Sioux as a Mathware Engineer in the Sioux Mathware Competence Center (formerly known as LIME). Within the Mathware department, Marloes strengthened her portfolio with projects related to data analysis, blockchains, neural networks and more. Note: After subscription, showing up is mandatory. Unsubscribing is possible until Monday the 2nd of March, 23.59h

This lunch lecture will be opened by Dr. Yanja Dajsuren, director of the PDEng Software Technology (ST) program at TU/e. She will introduce briefly the PDEng programs at TU/e and then welcome Ani Megerdoumian as a speaker for the lunch lecture. Ani Megerdoumian is the winner of the PDEng ST Design Award 2020. She will talk about her PDEng graduation project at ASML with a focus on AI and Software Design. An abstract of her project can be found below. There will be a Q&A session at the end of the lunch lecture. As the complexity of any system grows, the need for diagnostics becomes essential. In this sense, the data produced as an input for any complex system, like the high-tech machine of ASML is critical. During the wafer production cycle in TWINSCAN machines, the measurement, modeling, and applied corrections are all logged to a diagnostic file called MDL (Machine Diagnostic Log). MDL contains essential data about the behavior and performance of ASML’s machines, and this data helps designers (as well as support engineers) to understand the behavior of the machine during the production. Over the years, customers have started using informal MDL data as well. By design, MDL does not protect itself from incompatible changes. This project is initiated to analyze the possibility of converting all needed data to official XML-based files at low cost. As a roadmap, the current intention is to provide formal data to the customers. This, however, comes at a high cost if conversion is done manually. As a result of my PDEng graduation project, a support tool, with mappings of the log file content to XML-based file is delivered. Besides, an iterable pipeline and corresponding prototype for producing the mapping based on domain expert opinions is delivered. Finally, a prototype which provides an XML-based report for aiding the human resources in designing the XML-based formal reports is submitted. My PDEng thesis report presents the domain analysis done to model the major requirements needed to be transferred to the formal file format. In other words, transferring the meta information and place it in the XML-based reports to cover the needs of diagnostics architects and engineers. In the next phase of analysis and to decrease the costs of the transferring the data to the formal format, artificial intelligence methods are used to determine how many reports should be generated for each task in the machine and what information each report shall cover. Curious to know what Ani's project and/or a PDEng is about? Subscribe below!

Dear members, unfortunately this lunchlecture is in Dutch. Of je nu boodschappen doet, Netflix kijkt, een productvergelijking doet op Independer of met Google Maps naar een bestemming rijdt: het maakt bijna niet meer uit wat je doet of je laat data achter. En met die data wordt van alles gedaan, waarbij vaak de nadruk ligt op het voorspellen van gedrag. Zeker ook binnen de financiële branche. Machine learning is aan een opmars binnen de financiële dienstverlening bezig. Sterker: het summum van machine learning zien we in onze branche, in de beleggerswereld om precies te zijn. Grote beleggingsbeslissingen worden vaak niet meer genomen door mensen, maar door razendsnelle computers die alle beschikbare data continu combineren tot een ‘koop’ / ‘verkoop’ beslissing. Maar het kan ook veel laagdrempeliger ingezet worden. Credit checks bij het aanvragen van creditcards gaan voor het leeuwendeel geautomatiseerd, op basis van de achtergrondvariabelen van de aanvrager. Fraude-detectie bij schadeclaims wordt bij verzekeraars ondersteund door machine learning modellen die de waarschijnlijkheid van fraude per claim kunnen bepalen. Next Best Action modellen ondersteunen besluiten over de wijze en het moment van klantbenadering. Ook binnen Interpolis gebruiken we machine learning om van alles te voorspellen, van klantgedrag tot schades. Een aansprekend voorbeeld hiervan is de InbraakBarometer in de Interpolis SlimThuis app, die in 2019 meerdere keren in de media is geweest. In onze workshop nemen we je mee in de werking van deze Barometer en hoe deze tot stand is gekomen.

Meer informatie volgt snel!