Dynamically correlating low-cost sensors with Dylos DC1100 Pro

3. Prepare data

3.1 Dylos Pro DC1100

DC1100 made by Dylos, a branch name for carpet cleaner and vacuum, is powered with a 9V-wall plug. A recent update of DC1100 is DC1700 with battery and EMF shield. A "Pro" version indicated for the lowest size of the particle is 0.25µm and a normal version is duller with 0.5µm as the lower end. The maker limits the product's use to "detect levels of airborne particulates" and not to "health impacts for any given individual". To correlate the particle counts to the health impacts, I need to refer different sources.

DC1100
Display on DC1100 (taken in May 25, 2019, Hanoi) and a chart showing small particle count to air quality

I will work through 5 sources to correlate small particle counts with the concencentration of fine particle (PM2.5).

  1. Calculate PM2.5 by dividing mass of particles to the volume of air went through
  2. Use fitting coefficients from www.myhealthbeijing.com
  3. By a simple approach by billpentz.com
  4. Use fitting coefficients by www.aqmd.gov using a FEM GRIMM EDM180.
  5. Use fitting coefficients using a FEM MetOne BAM 1020
On rayhe.net listed an additional fittings but no coefficients published through smartairfilters.com.

3.2 Concentration by density and particle's volume

Assumed that the particle density (ρ) is 1.65E+12 µg/m3 and for the fine particle collection, the representative radius is 0.44µm. Then the mass of one particle is:

$m=\rho\times(\frac{4}{3}\pi\text(r)^3)$ $=(1.65\times10^{12}(\frac{\mu g}{\text(m^3)}))\times(\frac{4}{3}\pi(0.44\times10^{-6}\text(m^3))$ $=(5.89\times10^{-7})\text(\mu g)$

DC1100 displays the small particle on the left and the large one on the right per 0.01 ft3. To get the number of particles lower than 2.5µm in radius, and convert to per m3

$\text{#p=(small-large)}\times3531(\frac{0.01ft^3}{m^3})$

Finally, the mass of fine particles (PM2.5) per a volume of air can be calculated by:

$\text{PM}_{2.5}\text{ = #p}\times\text{m}$ $\text{ = (small-large)}\times3531\times5.89\times10^{-7}$ $\text{=(small-large)}\times2.08\times10^{-3}(\frac{\mu g}{m^3})$

Check the source of assumptions:

  1. Density of particles: 1.65 , 1.53 or 1.69 µg/m3
  2. Radius of a particles: mentioned in Air quality sensor network for Philadenphia which cited a study of seasonal size distribution in Seoul, Korea where 0.44 µm is the size of firewood fire's particles
3.3 Concentration by fittings in Beijing and back-calculate to PM2.5

On myhealthbeijing.com listed a speadsheet correlating Dylos raw readings with a AQI. Additional information on using DC1100 as a PM2.5 the fitting monitor are linked here. The fitting are expressed by the below equation:

$\text{AQI}_{US}\text{ = }3.31\times10^{-22}\text{x}^5 - 1.04\times10^{-16}\text{x}^4$ $+1.19.10^{-11}\text{x}^3 - 5.85 10^{-07}\text{x}^2 + 0.016\text{x}+9.43$

with goodness-of-fitting r2 = 0.999. Using the breakpoints given by US EPA, PM2.5 can be back-calculated.

3.4 Concentration by simple estimation

The author on this site claimed that the estimation was obtained from Dylos Corporation with a link to DC1100's patent. There is no other source to confirmed this approach. I emailed the custommer service at Dylos inquirying the PM2.5 to AQI conversion to no avail. This simple estimation is:

$\text{PM}_{2.5}\text{ = (small-large)/100}(\frac{\mu g}{m^3})$
3.5 Concentration by a FEM GRIMM EDM 180 fittings

The following equations are obtained by AQ-SPEC, an entity belonged to US EPA that done extensively testing with low-cost PM2.5 monitors.

An fitting equation between counting particle between 0.5-2.5µm (x) with DC1100 Pro version and an FEM monitor GRIMM EDM-180 with r2=0.815.

$\text{PM}_{2.5}=-8\times10^{-12}\text{x}^2+5\times10^{-05}\text{x}+3.98$
3.6 Concentration by a FEM BAM 1020 fittings

Similar to the approach with EDM-180, the following equation is the fittings with another FEM monitor, MetOne BAM-1020 with r2=0.632.

$\text{PM}_{2.5}=-1\times10^{-11}\text{x}^2+4\times10^{-05}\text{x}+4.17$