Pump Failure and Spill - Lessons Learned

Pump Failure and Spill - Lessons Learned

Hey all. So yesterday I was running an eVOLVER experiment overnight. The next morning, I received a text from a colleague in lab. “Hey Santi! Apparently the evolver overflowed (?) last night. Facilities came by and wants you to give them a call.” I came into lab and saw a completely empty bottle of media, some absorbent pads on the floor in front of our eVOLVER, and a small puddle on the bench underneath the unit. Notably, it was still running. After cleaning up, inspecting everything (pumps, lines, sleeves, etc.), and pulling the OD and temperature data, here are my observations:

Vial 3 of our unit overflowed and dumped an entire 1L bottle of media onto the top of the machine spill tray.

Efflux pump 3 could not move any liquid when tested. Upon closer inspection, a fine white dust was seen accumulated around the top of the pump, the metal shaft was turning, but the plastic rollers (white colored, like the aforementioned dust) were not. When disassembled, one of the plastic rollers in Efflux pump 3 showed scoring and abrasions, likely indicating that the metal shaft had been grinding against it. After cleaning and reassembly, the pump appeared to be working adequately (rollers were spinning, liquid was moving).

Sleeve 3 had media at the bottom of the sleeve.

The vacuum line to the spill tray was not turned on, I had forgotten the day before.

The OD and temperature traces for the vial that overflowed (vial 3) both show large aberrations occurring at the same time, likely due to media in the sleeve and therefore indicating the actual time of overflow. Interestingly, the OD reading begins to show extreme noise prior to the overflow event itself.

So, what happened here? This is what I think:

The cultures were growing normally (as seen in the other vial’s trace) and vial 3 successfully completed 1 dilution event at about hour 7.75. After this dilution event, vial 3’s OD reading began to suffer from increased noise. At hour 9.4, the OD reading began to show extreme noise (the reason for which I’m still unsure, any suggestions are welcome). This extreme noise eventually triggered a dilution event at hour 11.25. At this point, efflux pump 3 was not working. The vial began to overflow into the sleeve. The culture now filling sleeve 3 occluded the OD sensor, leading to an even higher OD reading, and the continued pumping of media into the vial until the bottle was empty. Eventually, the media drained out of the sleeve, revealing the OD sensor, and leading to the drop in measured OD at hour 14.5. It was only at this point, I believe, that influx pump 3 stopped running.

Moving forward I will certainly remember to turn on the vacuum line in case of something like this, and only a media bottle of sufficient volume to get through the night, but I definitely do not have much confidence in these pumps. The lack of a true mechanical connection between the shaft and the lobes (relying on only friction to actually make the pump function), makes it seem as if these pumps could fail at any time without warning. While I’m aware of the trick to put adhesive on the shaft to make it grip the rollers better, perhaps future iterations of the eVOLVER platform could benefit from more robust pumps. Additionally, perhaps some creative software safeguards could be implemented to prevent this kind of ‘overflow --> higher OD reading --> continual overflow’ cycle in the event of efflux pump failure (maybe using the huge temperature spike observed, or based on some kind of timing safeguard).

Anything I missed? Looking forward to hearing people’s thoughts. Thanks.

1 Like

Hi Santi. A bummer this happened. A few questions/ comments:

  • Adding a stop for high temperature is certainly something we should build in. We’ve been meaning to but theres been a bunch of changes from our end on the software. @heinsz @mgalardini this should be pretty easy to do right?
  • Are any of the other vials not working? Did the media get into the box itself or does everything appear to still be fine?
  • You will probably want to rebuild the sleeve that overflowed. Message me and I can send you some parts to help you rebuild it (e.g. new Vial Board, thermistor, photodiode, LED) Here’s a tutorial on how to build it.

I added a few more tips on how to apply the adhesive on the previous thread.

@santibk, overflows suck! Sorry about that!

@cmancuso, @mgalardini, and I have talked about putting in some sort of overflow detection based on weird data. A couple of things that could happen is your temperature curve goes extremely high, as in this case, or it could become extremely noisy, similar to right before it jumps up in the the plots @santibk shared. It’s definitely something we should consider implementing in the near future, possibly as a text/email alert.

My other thoughts are about when the efflux line stopped working. Assuming that was a 25 ml culture, and the desired dilution was from 0.4 to 0.2 (50% dilution, 12.5 ml media per dilution), it would take about two dilutions of no efflux to overflow the vial as the vials can only contain ~40 ml. Even if the dilution were only 25% (6.25 ml), two dilutions would put you right at the top of the vial or potentially over if the efflux straw is a little short. So most likely this efflux wasn’t working for the entire experiment. We always try to check that they are properly diluting media before we start the experiment, but sometimes things like this still happen.

@heinsz minor point but those volume calculations are off, remember that since its it’s continously diluting you need to calculate differently.

In turbidostat you can calculate how much is used with the following equation:
volume* ln⁡((upper thresh)/(lower thresh)) = media used per dilution
So for a 25mL culture held between OD 0.2 and 0.4, with a 2 hour doubling time, it’ll use ~17mL per dilution.

So overflow could happen during the first dilution. Smaller dilutions or smaller culture volumes would change that math though.
Agree with what others have said, let’s implement an overflow check in the next dpu iteration.

1 Like

@cmancuso Thanks for the clarification!