July 6, 2015 by Kevin Keras
A Positive Image
Seeing is believing, as the old saying goes. However, what separates laboratory automation from industrial assembly or manufacturing applications is its reliance on the assumption that everything will behave as taught. Well, as life teaches us...things change. Industrial automation has practiced total process control for decades. TPC aims to identify all variables in a process and predict the likelihood of failures. Only when you can identify and estimate the risks points in a process, can you predict (or control) the outcome.
Let's look at a common process involving a liquid handling robot performing a basic 96-well mtp plate replication. We'll keep it simple to illustrate the concept of TPC. Assume there is a nine position deck (3 x3) and the source plate (mother) is in location A1 (rear left) , the destination plate (daughter) is in A3 (rear right) and a box of disposable pipette tips is located in C1 (front left). Pretty simple method for the robot; Get tips from C1, aspirate 100ul from A1, dispense 100ul into A3 then eject tips back into A3. What could go wrong? A lot. First of all, we will assume that every location has been properly taught. Somehow, the right consumables have to be placed on the corresponding deck locators. This could be done manually by an operator or automatically by the liquid handling robot gripper or an external robot arm. Next, the robot the has to properly attach the tips and reliably aspirate/dispense the liquids, then eject the tips back into the tip box.
How do we know if the operator put the plates and tips in the right locations? How do we know that the plates or tips are oriented correctly or that they are seated flush in the locators? We don't..and neither do most liquid handlers. Some clever programmers have gone so far as to use visual prompts (pictures of how the deck should look) to assist operators in populating the deck prior to running assays. But, who is to say that the operator really gets it right? Additionally, what happens if a tip gets hung up on a mandrel? Crashes and spoiled assays can be the end result of 'things going wrong.' That is why process control is important. Now, what could we do to ensure proper consumable placement and operation/
Vision sensors are an ideal tool for quality control. Available from numerous vendors (Cognex, Keyence, Omron_ and others), vision sensors are compact and low cost devicess that allow users to capture a known good image which can then be used as a template to test for matches. They include built in cameras, illumination and I/O along with software for inspecting and training. Unlike Vision Systems, which are often highly programmable frame grabbers, sensors are go/no-go devices that provide fast 'bad/good' decision making.
A sensor could be attached to a liquid handling robot arm and driven over consumables prior to a run to see if the plate or tips match a template. Most sensors have the ability to save multiple images and can communicate via simple digital I/O. A robot developer can then use the robots I/O to initiate inspections that either verify and initiate a method or flag an operator with an error message. In my prior blog we talked about static issues and how they can impact disposable pipette tips. Imagine now, using vision sensor (looking horizontally) to inspect the robot mandrel for the presence of a hanging tip.
I am sure there are other possible uses for vision sensors in your lab operations...it wouldn't require a 'visionary' to come up with a few. Let us all know if you have some ideas?