A Machine Learning Approach for a Manufacturing Database of Nb-Ti Superconducting Wire

Abstract

This paper describes a machine learning approach for a manufacturing database in a Nb-Ti superconducting wire domain. A Nb-Ti superconducting wire is produced by iterating the drawing and heat treatment operations. A problem to determine a production schedule, or the parameters in the drawing and heat treatment operations, is discussed. We first show a database approach that selects the most promising schedule to achieve the target superconducting quality from the data. This aproach is useful in the effective reuse of the past experiences, but does not yield a novel production schedule Next we present a machine learning approach to extract knowledge to improve a production schedule for better superconducting quality. We use ID3 method to generalize the data, and we describe two criteria, correctness and applicability indices, to measure the quality of the induced knowledge.