A data mining and machine learning software for accelerated high-temperature corrosion-resistant materials design

Deposit-induced degradation of alloys and ceramic coatings in turbine engines is regarded as one of the core challenges preventing further gains in performance and efficiency. Emerging evidence suggests that synergistic effects in complex (mixed) deposits containing sulfates and silicate mixtures threaten material durability outside of classically-defined temperature ranges of CMAS degradation and hot corrosion (HC). These phenomena extend beyond the existing mechanistic understanding of the degradation processes, thus a new materials design paradigm capturing the effects of complex deposit chemistries and the gaseous environment is urgently needed to accelerate the implementation of next-generation materials. The proposed research addresses this need by developing a software tool employing machine learning (ML) to integrate insights from literature data, new experiments, and computational models to capture key CMAS and HC degradation (CMAS-HC) phenomena. The effort is led by QuesTek Innovations LLC and STTR RI partner Prof. David Poerschke at University of Minnesota. Prof. Carlos Levi and Prof. Brian Gleeson will be subject matter expert consultants. The team proposes to develop automated data retrieval & assessment toolkits that are able to handle multiple data formats and data sources. These toolkits will then be applied to develop comprehensive CMAS-HC databases with content diversity and format consistency. A ML-incorporated data analytical program enhanced by domain-knowledge will be constructed and implemented on the CMAS-HC databases to model materials’ resistance to CMAS-HC. Exploratory experiments to train and validate ML models will also be performed on mixed deposit systems. The ultimate goal is obtain improved understanding of complex mixed deposit CMAS-HC phenomena from ML and experiments, and leverage that understanding to enable mechanism differentiation and novel materials design