AMA 2019 Speakers
Dr. Anssi Laukkanen (Plenary Speaker)
VTT Technical Research Centre of Finland Ltd, Finland
Biography: Dr. Anssi Laukkanen is a Principal Scientist at VTT Technical Research Centre of Finland Ltd., responsible for development of Integrated Computational Materials Engineering solutions employing multiscale and multiphysics modeling. At VTT he is the responsible Principal Investigator for computational material sciences and engineering and the associated strategic scientific spearhead and leading the affiliated research group activities. He acts as the lead developer of the VTT properTune multiscale materials modeling solution and software toolset. His research interests include development of multiscale modeling techniques especially in the micromechanical range, consisting of modeling of single and polycrystal scale phenomena affiliated with deformation and failure behavior of materials. This includes development of hierarchically coupled and concurrent across-scales modeling solutions, focusing on metallic materials, metal based composites and thin films and coatings. Recent efforts have included material discovery and informatics related data centric methodologies and new computing paradigms, including emerging machine learning and artificial intelligence techniques for materials sciences and engineering.
Speech Title : Machine Learning Driven Design of Coatings to Combat Erosion Wear in Wind Turbine Blades
Abstract: Erosion of wind turbine blades (WTBs) by droplets, hails and other particles like sand is a prevalent problem in the wind power industry. Blade leading edge erosion influences negatively the power output of wind turbines and can lead to costly maintenance and repair operations let alone unscheduled downtime. The design of wear resistant coatings to meet the challenges of ever increasing blade tip speeds and extreme operating environments has proven demanding, and the wear problem still persists. The contribution of current work is in addressing coating design for WTBs by utilizing a machine learning (ML) exploiting design workflow. The idea of the concept relies in merging physical modeling, characterization and experimental results to data-driven ML models to enable the bridging of causal relations from fundamental material characteristics of, for example, a composite material to its performance dominating operating environment. Virtual models are used parallel to experimental results to provide training data to ML models focusing on characteristics such as composite structure, material properties and erosion wear performance in specific wear inducing conditions. The ML models provide a basis for a fast-to-compute high throughput workflow. This enables one to perform thorough optimization and material discovery of coating material solutions to mitigate WTB erosion damage based on the extensive design space of a specific coating material type. Use cases are presented to demonstrate the approach in practice in differing wear environments, based on different fitness metrics and types of wear resistant coating. The results present how data-driven modeling can contribute to wear resistant material design by enriching the capabilities of experimental methods and physics-based modeling workflows in yielding novel and improved material solutions.