S-2-27 A Screening Strategy for IN718 Combining High-throughput Forging Experiment and Machine Learning

A Screening Strategy for IN718 Combining High-throughput Forging Experiment and Machine Learning

Zhiron Sun, Kaikun Wang*

University of Science and Technology Beijing, Beijing, 100083

 

ABSTRACT: Nickel based superalloy IN718 is widely used in power station, pressure vessels, and other high-temperature structural applications. In this study, we propose a screen strategy of processing conditions based on high-throughput experiment equipment, numerical simulation, and machine learning to obtain the optimal process condition for IN718. A high-throughput experiment equipment for hot forging was designed, with which up to 30 samples of different processing conditions could be forged at a stroke. Numerical simulation of the forging process on the equipment was studied. Four factors of forging speed, forging displacement, friction factor and forging temperature were considered. Subsequently, A database of 625 examples with four features (forging speed, forging displacement, friction factor, and forging temperature) and two target properties (average grain size and maximum principal stress) were obtained. The database was splitted into 80% training data and 20% validation data. Several machine learning algorithms in skit-learn, a Python framework for data mining and data analysis, were compared based on training data. A BP NN model with minimum RMSE value was selected to analyze the influence of different processing conditions on average grain size and maximum principal size repectively. The analyzing results were consistent with the domain knowledge. A screen algorithm based on Python was designed to screen the optimal processing conditions comprehensively considering the average grain size and the maximum principal stress. The optimal conditions for different forging displacements were achieved. Compared with the traditional high-cost and time-consuming trial-and-error methods, the method used in this paper to optimize the processing technology has great advantages. This method could also be applied to pre-screening for material design and process optimization.

 

Keywords: Hot forging; Numerical simulation; Machine learning; High-throughput experiment

Brief Introduction of Speaker
Kaikun Wang

Prof.Dr-Ing Kaikun Wang: He works in School of Materials Science and Engineering at University of Science and Technology Beijing. He did his PhD work in Institute of Metal Forming at RWTH-Aachen University and his postdoc work at Tsinghua University. His main research interests include preparation and short-term forming technology of lightweight alloys and composites, superalloy forming technology, simulation of metal plastic forming process, high-throughput forming, etc.