A method to optimize sampling locations for measuring indoor air distributions

Yan Huang, Xiong Shen, Jianmin Li, Bingye Li, Ran Duan, Chao Hsin Lin, Junjie Liu, Qingyan Chen

Research output: Journal article publicationJournal articleAcademic researchpeer-review

14 Citations (Scopus)

Abstract

Indoor air distributions, such as the distributions of air temperature, air velocity, and contaminant concentrations, are very important to occupants' health and comfort in enclosed spaces. When point data is collected for interpolation to form field distributions, the sampling locations (the locations of the point sensors) have a significant effect on time invested, labor costs and measuring accuracy on field interpolation. This investigation compared two different sampling methods: the grid method and the gradient-based method, for determining sampling locations. The two methods were applied to obtain point air parameter data in an office room and in a section of an economy-class aircraft cabin. The point data obtained was then interpolated to form field distributions by the ordinary Kriging method. Our error analysis shows that the gradient-based sampling method has 32.6% smaller error of interpolation than the grid sampling method. We acquired the function between the interpolation errors and the sampling size (the number of sampling points). According to the function, the sampling size has an optimal value and the maximum sampling size can be determined by the sensor and system errors. This study recommends the gradient-based sampling method for measuring indoor air distributions.

Original languageEnglish
Pages (from-to)355-365
Number of pages11
JournalAtmospheric Environment
Volume102
DOIs
Publication statusPublished - 1 Feb 2015

Keywords

  • CFD simulation
  • Error analysis
  • Gradient method
  • Kriging interpolation

ASJC Scopus subject areas

  • Environmental Science(all)
  • Atmospheric Science

Fingerprint

Dive into the research topics of 'A method to optimize sampling locations for measuring indoor air distributions'. Together they form a unique fingerprint.

Cite this