search

RespiraWorks / Ventilator

Fully-featured ICU ventilator design, optimized for manufacture using commonly available components and free to license. Repository tracks all mechanical, electrical and systems design, software, requirements and regulatory documentation.
Apache License 2.0
130 stars 37 forks source link

Venturi dP characterization coefficient #631

Open esc-works opened 4 years ago

esc-works commented 4 years ago

One issue with the reliability of the venturis is that we are using an "ideal" response coefficient, but we are working in the physical work. @sglow has done some calibration with an air cylinder, and found the response was reliably 12% low, flow

We have done some testing with the pneumotachograph, and looks like the flow was about 12% high. image (26)

Both look highly linear though, so I think we need to do better about coming up with a characterization curve since it probably has to do with manufacturing variation and consistency in how we are doing the calculation.

  1. measure the throat and channel diameters with a micrometer and pin guage to remove any variation in the calculation of throat diameters.
  2. Calibrate a known diameter venturi against the pneumotachograph and develop a linear offset correction factor.
  3. Try printing an additional venturi with slightly different throat diameter to check scalability of correction factor.
esc-works commented 4 years ago

Datasheet for AMS5915-0100-D differential pressure sensor. ams5915-5.pdf I2C-Connection-Diagram-105.pdf

esc-works commented 4 years ago

@brmurray we applied a linear correction factor the flowrate, but we actually should have applied a linear correction factor to dP. Something like Q_correct = Q_read A Sqrt(dP)

esc-works commented 4 years ago

We ran a much better set of testing. Using the SLA printed venturis and comparing against a Fleisch pneumotachograph, we used a linear correction factor 0.97 based on the discussion here. https://www.lmnoeng.com/venturi.php

The throat value is 5.5 mm. The pipe value is 15.05 mm.

We also compared using a more expensive i2c differential pressure sensor which showed no change. I'm tempted to call this good until we acquire better data.

The plot below shows Flow recorded by the pneumotachograph (PTG) in L/s. The red points are computed by calculating a flow rate from recorded differential pressure using a digital AMS i2c sensor. (record pressure, calculate flow, hence the perfect quadratic fit). The blue points are same but using a different analog pressure sensor from NXP. (the agreement between the two sensors was very good). Note that the third fit line is labeled PTG but it should be NXP.

image (28)

inceptionev commented 4 years ago

making a comment here to link this ticket to the archived slack channel #calibration-tests where this some of this discussion took place: https://respiraworks.slack.com/archives/C014BB987TQ

martukas commented 3 years ago

Venturi geometry has changed and we probably need to redo this. Furthermore, we probably want to come up with a reproducible protocol that can be easily learned and used by people not intimately familiar with this. This would go with using the Ventmon.