Trosifol: “SoundLab AI” tool can predict sound insulation values

Case study develops prediction tool to derive sound insulation values of arbitrary glass assemblies

In a recently presented joint case study, Michael Drass, Michael Anton Kraus, Henrik Riedel (all M&M Network-Ing) and Ingo Stelzer, Kuraray Europe GmbH, developed an AI-based software tool for deriving sound insulation values of arbitrary glass assemblies.

The idea was to predict the weighted sound insulation value for glazing systems, as this value can only be determined by very complex numerical simulations or expensive experiments in classical approaches. The presented ML tool was trained on structured data in a supervised learning procedure. The data were obtained within an extensive experimental program. The accuracy in the prediction error plot shows a very high predictive ability, which could be proven by an R2 = 0.996 for training data and R2 = 0.982 for validation data. In addition, the ML model was also checked for previously unexploited test data.

The authors concluded that the developed tool is a suitable method to make predictions about the sound insulation of arbitrary glass structures quickly, cost-effectively and efficiently, which is a great advantage for the designing architects and engineers, especially in early project phases.

The detailed case study and the corresponding tool, the app, is offered on the website of Kuraray Europe Gmbh and can be downloaded here.