Testing Protective-Mask Materials

[lpetrich]

What Are The Best Materials for Making DIY Masks? - Smart Air Filters
notes
(PDF) Testing the Efficacy of Homemade Masks: Would They Protect in an Influenza Pandemic? - some work back in 2013

A surgical mask performed very well, with only a vacuum cleaner bag being comparable. But it was much more difficult to breathe through a VCB than through a SM. A cotton T-shirt captured 2/3 of 1-micron particles and 1/2 of 0.02-micron particles -- and was like the SM in breathability.

The Ultimate Guide to Homemade Face Masks for Coronavirus - Smart Air Filters

Expands on that work with some homemade test equipment - "Henderson apparatus". The experimenters did tests on 30 different materials, 3 times how many the Cambridge ones tested -- and likely much more representative.

Results: Ebola-Sized Particles (1 micron)

For Ebola-sized particles, the N95 mask, surgical mask, and HEPA filter performed best, capturing over 99% of particles 1.0 micron and above. Next up, the HERO coffee filter captured 98%. Paper towels, canvas, denim, and the cotton bed sheet also captured more than 90%.

All materials were far better than nothing; most blocked over 50% of particles. But the four worst-scoring materials were the wool scarf, polyester neck warmer, cotton bandana, and light scarf.

Results: Smallpox-Sized Particles (0.3 microns)

For 0.3-micron particles, there was a much wider range in effectiveness. The N95 mask, HEPA filter, and surgical mask still did best, all capturing over 75%. However, the materials consistently captured fewer smaller particles than larger particles.

Among the household materials, the HERO coffee filter came up next in the list, capturing 62%. But only four other materials filtered more than 48%: the 40D nylon, CHEMEX coffee filter, the dish towel, and canvas.

Some materials were only slightly better than nothing. The bandana, neck warmer, scarves, cleaning cloths, and 100% cotton T-shirt all captured less than 10%.

Scarves were among the worst. Natural fibers did better than synthetic ones, because they are rougher, meaning that they are better at capturing stuff.

 

[Valjean]

Thanx, interesting article.

Even the best material won't do well if it doesn't fit. I see masks all the time being worn with big gaps around the nose and the cheek sides pooched out. What really baffles me are the masks being worn only over the mouth.
What the heck?!

 

[lpetrich]

A big problem with some of the materials tested is that they are difficult to breathe through. So the testers constructed a test of that: find out how much pressure is necessary to keep some amount of airflow through each material. Too much means too difficult to breathe through.

Not surprisingly, the best materials for stopping particles were also the worst for breathing through.

The double layer 100% cotton T-shirt, bra pad, 70D nylon, paper hand drying towel, cotton bed sheet (120 thread) and denim (10oz) all fell between the breathability of the surgical mask and N95 mask.

Then a lighting test -- one which wasn't very good, it seemed to me. Their recommended materials:

• Denim (10oz)
• Bed sheets (80-120 thread)
• Paper towel
• Canvas (0.4-0.5mm thick)
• Shop towels

P.S. What is 10 ounce denim anyway?

Denim is typically measured in weight per square ounce. A square yard of 10oz denim would weigh 10oz. Anything under 12oz is considered ‘lightweight’ for denim pants, so you can think of the 10oz denim we tested as being similar to that from lightweight jeans.

I've downloaded the image files and I'll be OCRing them to get their numbers. I'll then use Mathematica to compare the different measurements to each other.

 

[lpetrich]

I couldn't find any algorithm-complexity options for Mathematica's Predict[] function, because with enough parameters, one can fit *anything*.

Here are my results. I used the function's default algorithm settings, like automatic choice of algorithm.

For predicting 0.3 um from 1.0 um results:

• Algorithm: Nearest Neighbors
• Trained on all: 5.95
• Runs: 100
• Trained on 1/2: 7.28 +- 5.29
• Tested on rmdr: 13.73 +- 4.55

For predicting breath from 1.0 um results:

• Algorithm: Linear Regression
• Trained on all: 1.38
• Runs: 100
• Trained on 1/2: 1.29 +- 0.35
• Tested on rmdr: 1.52 +- 0.28