Several temperature sensors were placed around the refrigerated well to check it's baseline performance with a very simple temperature control algorithm.
- T0 is on the bottom of the well, between the blood products and the bare metal of the well. This is the coldest location in the well.
- T1 is on the coolant tube, where it enters the well. This is the coldest location on the entire cart.
- T2 is on the side wall of the well, between the blood products and the bare metal of the well. This is the second coldest location in the well.
- T3 is on top of the blood products. This is the warmest location in the well.
- IR sensor is in the center of the lid, looking straight down onto the blood products. It has a wide field of view that likely includes some of the side wall.
Temperature Control Algorithm
- Input from T0
- Above 1.5 C, refrigeration switches on
- Below 0.75 C, refrigeration was switches off
- 12 RBC units
- 12 FFP units
T3 is above the limit of 6C. The difference between the coldest place (T0) and the warmest place (T3) is between 6 and 10 C. I think improved insulation will reduce this difference. Insulation ideas include improved lid sealing, double pane lid, and moving the lid seal outward to include more of the well's lip.
T0 dips below 0 C for a minute or two (less than 25% of the time) each time the refrigerator is activated. This brief dip in temperature does not seem to be transmitted to the blood products (they do not freeze, both T2 and IR readings are well above freezing) however it would be product to check for hemolysis before loading viable units.
In production carts, it is desirable to eliminate contact-type temperature sensors in the refrigerated well because they are prone to misplacement and breakage. Therefore this testing includes non-contact type sensors (the IR sensor) positioned to measure the warmest part of the cargo (the top), and a thermocouple positioned outside the well at the coldest point overall. As more data is accumulated, correlations will be developed between direct contact measurements and remote readings to allow them to be used for temperature control.
All of the sensors are read several times per second and analyzed by the control computer.
Data is periodically written to a log file at a lower frequency, and augmented by data points when significant events occur (e.g. the temperature exceeds a limit). This strategy seems to capture reasonable picture of the cart's activity while keeping the log file sizes manageable. Last night's 16 hour run generated 354 kb of data (5,700 data records). Extrapolating, a 24 hour day would generate 0.5 megabytes (8,550 records). At this rate, the cart could run continuously for 22 years before filling up a 4 Gb SD card (it's hard to find smaller cards these days).