Stéphane Dauzere-Peres is Professor at the Center of Microelectronics in Provence (CMP) of Mines Saint-Etienne in France and Adjunct Professor at BI Norwegian Business School in Norway. He received the Ph.D. degree from the Paul Sabatier University in Toulouse, France, in 1992; and the H.D.R. from the Pierre and Marie Curie University, Paris, France, in 1998. He was a Postdoctoral Fellow at the M.I.T., U.S.A., in 1992 and 1993, and Research Scientist at Erasmus University Rotterdam, The Netherlands, in 1994. He has been Associate Professor and Professor from 1994 to 2004 at the Ecole des Mines de Nantes in France. He was invited Professor at the Norwegian School of Economics and Business Administration, Bergen, Norway, in 1999. His research interests broadly include modeling and optimization of operations at various decision levels (from real-time to strategic) in manufacturing and logistics, with a special emphasis on semiconductor manufacturing. He has published more than 65 papers in international journals. He has coordinated multiple academic and industrial research projects. He was runner-up in 2006 of the Franz Edelman Award Competition, and won the Best Applied Paper of the Winter Simulation Conference in 2013.
Talk # 1
Some Challenges on Integration of Decisions in Logistics
In this talk, we first present the notions of horizontal integration and vertical integration of decisions in logistics, with their main characteristics and motivations. Challenges related to each integration type are then discussed using examples based on academic and industrial research conducted by the author. The integration of decisions in production planning and scheduling and a railway transportation example are discussed for vertical integration. A production planning and vehicle routing problem and a maritime supply chain example are presented for horizontal integration.
Talk # 2
Achievements And Lessons Learned From A Long-Term Academic-Industrial Collaboration
I had the opportunity to work for about 14 years on many different projects with two manufacturing sites of the French-Italian semiconductor company STMicroelectronics. Supported by European, national and industrial projects, this still active long-term academic-industrial collaboration led to many scientific and industrial achievements, spreading to other companies. Through regular exchanges, engineers, researchers, PhD and Master students were able to present their problems, their advances and generate new research projects. After some history of the collaboration, the presentation will survey some of the main research and industrial results in qualification and flexibility management, production and capacity planning, scheduling, automated transportation, dynamic sampling and time constraint management. Challenges faced and lessons learned when applying Operations Research and Industrial Engineering in practice, and in particular in semiconductor manufacturing, will be discussed. Benefits for both practitioners and researchers will be emphasized, such as the opportunity to propose and study new relevant problems and develop and apply novel approaches using actual industrial data.
Tempe (AZ), USA
John Fowler is the Motorola Professor of Supply Chain Management in the W.P. Carey School of Business at Arizona State University. He served as the department chair of supply chain management from 2011-2016. Prior to that he was the Avnet Professor of Industrial Engineering at ASU. His research interests include discrete event simulation, deterministic scheduling, and multi-criteria decision making. He has published more than 120 journal articles and more than 100 conference papers. Professor Fowler recently served as editor-in-chief for IIE Transactions on Healthcare Systems Engineering and continues to be the department editor for Healthcare Operations Management at the journal. He is an editor of the Journal of Simulation and associate editor of IEEE Transactions on Semiconductor Manufacturing. He is a fellow of the Institute of Industrial and Systems Engineers and recently served as the vice president for continuing education. He is a former INFORMS vice president, and served on the Winter Simulation Conference board of directors from 2005-2017. He was also the program chair for the 2002 and 2008 Industrial Engineering Research Conferences and the 2008 Winter Simulation Conference.
Talk # 1
A Framework for Designing Remote Diagnostics Networks for Equipment Suppliers
With advances in information technology, service activities for expensive equipment used in semiconductor manufacturing can be performed from a remote location. This capability is called remote diagnostics (RD). RD has the potential to reduce maintenance and capital costs and improve productivity. In this presentation, we develop a queueing-location model to analyze the capacity and location problem of after sales service providers, considering the effects of RD technology. Our model optimizes the location, capacity and the type of service centers while taking congestion effects into consideration. We solve this model using a simulation optimization approach in which we use a genetic algorithm to search the solution space. We demonstrate how our methodology can be used in strategic investment planning regarding the adoption of RD technology and service center siting through a realistic case study.
Talk # 2
An Overview of Scheduling with Batch Processing Machines in Semiconductor Manufacturing
Batch processing machines play an important role in the operational performance of semiconductor manufacturing factories, both wafer fabs and assembly/test facilities. In this context, a batch is a set of lots that are processed at the same time on a single machine. Not only is there a trade-off between the time spent by lots waiting to form a batch versus the time lots spend waiting for a machine to become available (based on running partial loads), but the amount of variability introduced by various approaches to forming batches and scheduling them can make a large difference on the performance of these machines as well as downstream machines (and thus the factory performance). In this paper we will focus on situations where the time to process a batch is either a constant based on the recipe (so called incompatible families) or the maximum time required by any of the lots in the batch (so called compatible families). Together these are known as p-batch scheduling. The primary focus of paper will be on offline, deterministic models of semiconductor machine environments that include at least one batch processor (single machine, parallel machines, flowshops, and jobshops) and will discuss both the real world situations these models represent in semiconductor manufacturing and what solutions techniques are used.
Young Jae Jang received his Ph.D. degree in mechanical engineering from MIT in 2007 and a double M.S. degree in mechanical engineering and operations research from MIT in 2001. He received a B.S. degree in aerospace engineering from Boston University in 1997. He is currently an Associate Professor in the Industrial and Systems Engineering Department at KAIST, South Korea. His current research includes automated material handling systems (AMHS) design, and system design and analysis particularly for wireless power transfer–based logistics systems. He has been actively working with companies including Samsung Electronics, LG Electronics, and Samsung Heavy Industry. In early 2018, he received a USD$2 million research grant from Shinsung Factory Automation Ltd., one of the largest global AMHS solution providers in the semiconductor industry, and founded the Shinsung-KAIST AI AMHS Research Center at KAIST. As Director of the official KAIST research center, he is responsible for commercializing the new AMHS control system for semiconductor fabs using the learning-based vehicle routing and dispatching algorithms developed by his research group. Dr. Jang was the Co-Chair of the 2015 International Symposium on Semiconductor Manufacturing Intelligence (ISMI). He is currently the Guest Editor of a Special Issue on “Artificial Intelligence in Manufacturing and Logistics Systems: Algorithms, Applications, and Case Studies,” for International Journal of Production Research. Dr. Jang is also currently an Associate Editor of Computers & Industrial Engineering.
Talk # 1
Deep Q-Learning Based Semiconductor AMHS System Design and Industry Case Study
The learning based dynamic routing algorithm is proposed for the overhead hoist transport (OHT) system for semiconductor fabrication facility (FAB). The OHT system, which consists of multiple vehicles moving at high speeds on guided rails, is the primary automated material handling system (AMHS) in FABs. Modern large-scale FABs have hundreds of vehicles delivering lots between processing machines. The dynamic routing method is the route guidance method that dynamically selects the vehicles’ paths by considering the conditions of traffic and congestion. We develop a reinforcement learning-based dynamic routing algorithm called
QLBWR($\lambda$), which consists of a dynamic Boltzmann softmax policy and reward shaping on a Q($\lambda$) learning method. The proposed algorithm uses real-time information to effectively guide each vehicle so that it avoids congestion and finds its optimal path. The algorithm is also designed such that the computational burden to find its optimal route is significantly low enough to serve hundreds of vehicles in real time. The performance of the proposed algorithm is compared with various static and dynamic algorithms with simulation analyses on an actual FAB layout. The results show that the algorithm outperforms and is superior to the other benchmarking algorithms.
Talk # 2
Dynamic scheduling of the dual stocker system using reinforcement learning
The stocker system is the most widely used material handling system in LCD and flat panel fabrication facilities (FABs). The stocker mainly consists of one or two cranes moving along a single track to transport lots, or cassettes, containing 10 to 30 thin glass substrates between processing machines. Because the stocker system is the primary material handling system in the FABs, its performance directly affects the overall performance. In this study, we investigate the scheduling of a dual stocker system operating with two cranes simultaneously on a single track and propose a learning-based scheduling algorithm for the system. We report some of the results of our long-term efforts to dynamically optimize the dual-crane stocker. We show the modeling and algorithm to minimize the make-span of the jobs. In particular, we incorporate a reinforcement learning method in the scheduling algorithm. The model is validated in an extensive simulation study based on actual data.