Data Science Meets High-Tech Manufacturing – The BTW 2021 Data Science Challenge

For its third installment, the Data Science Challenge of the 19th symposium “Database Systems for Business, Technology and Web” (BTW) of the Gesellschaft für Informatik (GI) tackled the problem of predictive energy management in large production facilities. For the first time, this year’s challenge was organized as a cooperation between Technische Universität Dresden, GlobalFoundries, and ScaDS.AI Dresden/Leipzig. The Challenge’s participants were given real-world production and energy data from the semiconductor manufacturer GlobalFoundries and had to solve the problem of predicting the energy consumption for production equipment. The usage of real-world data gave the participants a hands-on experience of challenges in Big Data integration and analysis. After a leaderboard-based preselection round, the accepted participants presented their approach to an expert jury and audience in a hybrid format. In this article, we give an overview of the main points of the Data Science Challenge, like organization and problem description. Additionally, the winning team presents its solution.

Enhanced Failure Analysis (FA) of Organic Contamination Using Submicron Simultaneous IR and Raman Spectroscopy: Breakthrough Developments of Optical Photothermal IR (O-PTIR)

Rapid identification of organic contamination in the semi and semi related industry is a major concern for research and manufacturing. Organic contamination can affect a system or subsystem’s performance and cause premature failure of the product. As an example, in February 2019 the Taiwan Semiconductor Manufacturing Company (TMSC), a major semiconductor manufacturer, reported that a photoresist it used included a specific element which was abnormally treated, creating a foreign polymer in the photoresist resulting in an estimated loss of $550M [1].

Influence of relationship conflicts with leaders and coworkers on employees' voice

Building on self-determination theory and relational attribution theory, in this study we examined how relationship conflicts with leaders and coworkers simultaneously affect employee voice behaviors. We expanded relational attribution theory by developing two new constructs we labeled leader-relational attribution orientation and coworker-relational attribution orientation to describe employees' different responses to relationship conflicts with leaders and coworkers via psychological needs satisfaction. We surveyed 328 employee–leader dyads who were employed at a semiconductor manufacturer to test our hypotheses. We found that leader-relational attribution orientation can strengthen the influences of relationship conflicts with leaders on psychological needs satisfaction and its indirect effects on employee voice behaviors. Coworker-relational attribution orientation can strengthen the influences of relationship conflicts with coworkers on psychological needs satisfaction and its indirect effects on employee voice behaviors. Theoretical and practical implications are discussed.

Effect of Irrelevant Variables on Faulty Wafer Detection in Semiconductor Manufacturing

Machine learning has been applied successfully for faulty wafer detection tasks in semiconductor manufacturing. For the tasks, prediction models are built with prior data to predict the quality of future wafers as a function of their precedent process parameters and measurements. In real-world problems, it is common for the data to have a portion of input variables that are irrelevant to the prediction of an output variable. The inclusion of many irrelevant variables negatively affects the performance of prediction models. Typically, prediction models learned by different learning algorithms exhibit different sensitivities with regard to irrelevant variables. Algorithms with low sensitivities are preferred as a first trial for building prediction models, whereas a variable selection procedure is necessarily considered for highly sensitive algorithms. In this study, we investigate the effect of irrelevant variables on three well-known representative learning algorithms that can be applied to both classification and regression tasks: artificial neural network, decision tree (DT), and k-nearest neighbors (k-NN). We analyze the characteristics of these learning algorithms in the presence of irrelevant variables with different model complexity settings. An empirical analysis is performed using real-world datasets collected from a semiconductor manufacturer to examine how the number of irrelevant variables affects the behavior of prediction models trained with different learning algorithms and model complexity settings. The results indicate that the prediction accuracy of k-NN is highly degraded, whereas DT demonstrates the highest robustness in the presence of many irrelevant variables. In addition, a higher model complexity of learning algorithms leads to a higher sensitivity to irrelevant variables.