Automating membrane filtration in a food microbiology lab

In short

A food producer's microbiology lab tests sugar solutions for microbes. Each test pulls a measured amount of sample through a thin membrane filter under vacuum. The microbes stay on the membrane, the membrane is laid onto a dish of growth medium (an agar plate), and the plate goes into an incubator so that any microbes grow into countable colonies.

All of this is done by hand today, for up to 100 samples a day, and every handling step is a chance to contaminate the sample. Contamination shows up as a false result. The lab wants a system that runs the whole sequence, keeps it clean between samples, and records what happened to every sample.

How one sample is processed

Every step below is manual today. The system has to do all of them, repeating the filtration-to-plate part for every plate a sample needs.

Figure 1One sample, step by step· © Birdwave

Before a run

• Disinfect the robot system, possibly with UV.
• The operator loads sterile filtration racks, frits, funnels, agar plates, membrane filters and pipette tips.
• The operator provides sterile process water, a liquid waste container and waste bags.
• Prepared, sterile sample solutions come in through a defined hand-over point.

For every sample

1. 1
   Identify the sample. Pick it up and scan its QR code. The code selects the filtration protocol: sample volume, number of filtrations, target medium and incubation conditions.
2. 2
   Get into the sample vessel. Septum lid, pierceable membrane, or opening and re-closing the lid automatically. Not decided yet.

For every plate

1. 3
   Build the filtration unit. Take a membrane from the dispenser and place it on the frit, then pick up a sterile funnel and lock it onto the filter block.
2. 4
   Dose. 100 ml of sterile process water into the funnel. Take a fresh sterile pipette tip and dose the protocol's sample volume (3, 15 or 40 ml).
3. 5
   Filter. Open the vacuum valve. A pressure or liquid sensor detects the end of filtration, then the valve closes.
4. 6
   Wash. Another 100 ml of sterile process water, filtered again.
5. 7
   Plate it. Present the right agar plate and scan its QR code. Open it, lift the membrane off the frit, lay it on the agar, close the plate.
6. 8
   Stack. The plate goes onto the stack for its incubation temperature.

Figure 2The filtration station, in section· © Birdwave

After every sample

• Funnel: removed and stacked for cleaning and autoclaving.
• Frit: either swapped for a sterile one, or the whole filtration rack moves to a UV zone while a second sterile rack takes over.

What one sample turns into

Which scheme applies depends on whether the sugar solution was boiled. V1 to V4 are four different agar media.

Figure 3What one sample turns into· © Birdwave

Every sample volume goes on top of 100 ml of sterile water and is followed by a 100 ml wash, so up to about 240 ml passes through one funnel per filtration.

Keeping it sterile

Every plate is a test: whatever the system carries from one sample to the next grows on the agar and reads as a false result. The negative controls above exist to catch exactly that. So the process has to stay as sterile as possible from start to finish, with cleaning between every sample so nothing carries over.

To be evaluated:

• an enclosure with UV disinfection, and whether UV is enough on its own
• alternatives or additions: hydrogen peroxide (H₂O₂), infrared, ultrasound
• automatic disinfection cycles for the parts that touch every sample: grippers, frits, filtration units
• whether a manual ethanol wipe before production starts is still needed

When a filtration stalls

Rare with white sugar, but the system needs a plan for it. The lab proposes a 10-minute limit and automatic fault detection, then one of two recoveries:

• Variant A: break the filter. Alarm, break the filter so the vacuum pulls the liquid through, discard the filter, clean the frit, insert a sterile frit, carry on.
• Variant B: pipette it off. Alarm, pipette the remaining liquid off (automatically or by hand), discard the filter, clean the frit, swap in a sterile one, carry on.

Where the system sits

Figure 5Where the system sits· © Birdwave

Consumables and storage

Funnels, agar plates, membrane filters, pipette tips and frits are stored on board and presented automatically: a carousel, stack magazines or flexible stacking systems. Finished plates need buffered storage, kept apart by incubation temperature, until they go into the incubators.

Water supply and waste

• Supply: sterile process water in bag systems, with feed pumps and dosing.
• Waste: a liquid waste container, automatic disposal of solid waste, replaceable waste bags.

Interfaces

• Upstream, which may be automated later: dissolving the samples, preparing the media, pouring the agar plates.
• Downstream: moving finished plates into the incubators, recording the data, tracking every sample by QR code.

Traceability

Every step is recorded against the sample: sample number, filtration parameters, the agar plate used, timestamps and every operator intervention.

Decisions still open

Figure 6Decisions nobody has made yet· © Birdwave

Robot concept

• Variant A: a robot arm with a tool changer swapping between a gripper, suction, a pipetting module and membrane handling.
• Variant B: a gantry with fixed tools, moving in height and width. No tool changes, and more work can run in parallel.

Neither has been chosen, and the economics have not been compared.

Open technical questions

• Are larger robotic pipette tips available for volumes up to 40 ml?
• Can the frits be magnetic, to make them easier to handle? Can they be cleaned automatically?
• Does each vacuum position need to be controlled on its own?
• What is the best number of parallel filtration units: strips of 2, 5 or 6?
• How are samples handed over sterile from the upstream process?
• Single-use or reusable components: which is cheaper overall?

What the figures imply

• 200 to 500 filtrations, membranes and plates a day at 100 samples, plus the negative controls.
• About 40 to 110 litres of water and sample through the system and into waste each day.
• Filtrations must overlap. 5 minutes per sample for up to 5 filtrations, each with a vacuum step, a wash and a membrane transfer, cannot run one after another. That is why the number of positions per strip matters.
• About 24 samples on board. 120 minutes of walk-away at 5 minutes per sample means up to about 120 plates, membranes and tips, and 24 funnels and frits or more, loaded before the operator walks away.
• One long shift. 100 samples at 5 minutes each is about 8.3 hours.

What a useful proposal covers

A concept does not have to answer everything at once, but it should say how it handles:

• the station layout and robot variant, with a cycle-time estimate against 5 minutes per sample
• moving the membrane from frit to agar without folding, tearing or trapping air
• disinfection between samples, and how it would be validated
• how many consumables it holds, and how that reaches 120 minutes of walk-away
• dosing 3 to 40 ml within ±10 % without under-dosing
• stalled-filtration recovery and the traceability record
• single-use versus reusable parts, with running costs