An Inverse Design Method for Windshield Defrosting-Demisting Performance Using Machine Learning Techniques.

As vehicle electrification progresses, physical spaces inside instrumental panel for a defroster nozzle become narrower and nozzle sizes are required to reduced. Under such constraint conditions, it is important to design a defroster nozzle, satisfying requirements of windshield defrosting-demisting performance with low costs. In this paper, a reduced order model (ROM) is developed to predict instantly windshield velocity distribution, namely windshield defrosting-demisting performance. An inverse design method utilizing decision tree algorithm and ROM is established to find out instantly design conditions of defroster nozzle that fulfill performance requirements. Effectiveness of the method is validated by CFD performance results of a small defroster nozzle derived from the method results.

Language

  • English
  • Japanese

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01766281
  • Record Type: Publication
  • Source Agency: Japan Science and Technology Agency (JST)
  • Files: TRIS, JSTAGE
  • Created Date: Jan 25 2021 3:37PM