Valence-Bond Based Machine-Learning Structure-Activity Relationship Models to Assess the Performance of Gaseous Dielectrics in Multidimensional Aspects

In view of the growing impact of sulfur hexafluoride (SF<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</inf>) on global environment, the development of eco-friendly dielectric replacement gases has attracted considerable experimental and theoretical interests. A valence-bond based structure-activity relationship method is proposed in this work to assess the performance of a gas in six aspects including dielectric strength, boiling point, global warming potential, arc interruption capability, toxicity, and flammability. The chemical bonds are extracted from the simplified molecular input line entry specification (SMILES). Good correlations between theory and experiment are achieved using the extensive machine-learning training by means of the artificial neural network, shedding new light on the characteristic sought in SF<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</inf> replacement gases.