Challenge:

Develop an automated solution for the inspection of lyophilised beads used in diagnostic assay (an investigative analytic procedure) development in order to scale-up manufacturing for new markets while maintaining high accuracy repeatability. 

Background:

Biofortuna is a trusted outsourcing partner for the provision of laboratory services, custom diagnostic assay development and contract manufacturing. A key service  is the development and manufacture of customised lyophilised beads - a single, accurate dosage of reagent, a substance or compound that can facilitate a reaction most widely used in diagnostic tests, which is lyophilised (freeze dried) as a sphere.

The process of freeze drying heat-sensitive materials - such as microbes, proteins, pharmaceuticals, plasma and tissues - improves the shelf life and allows them to be transported without needing to use expensive cold storage and transport solutions. 

Biofortuna uses lyophilisation to freeze-dry reagents and chemicals used in the process of analysing DNA samples and determining the presence of diseases. The beads are typically about 20 microliters - equivalent to about three millimetres in diameter - but can range from five microliters up to 30 microliters. 

Biofortuna wants to encourage wider use of lyophilised beads across new sectors because of their environmental and accessibility benefits. They were signposted to AMRC Cymru by the Welsh Government to take advantage of its Accelerating Decarbonisation and Productivity through Technology and Skills (ADAPTS) programme. 

The company did not have an accurate way to assess or quantify the defect rate of lyophilised beads following the freeze drying part of the process. This prevented it from having clear data on the success rate of the manufacturing process at this stage in production. 

Typically, the trays of beads are photographed but defects are not counted or assessed, instead it is manually inspected by sieving beads onto a tray to capture any that are fragmented or misshapen. 

If no defects are found, it is passed into a semi-automated placement jig, which has slots to sort into respective vials on a 96 well plate.

Biofortuna manually assess the beads after the lyophilisation process for defects, and a second time before packaging for bead presence and identification of defects when the lyophilised beads are on the 96 well plate.

Even though most defected beads are removed through a sequence of sieves, this method is not 100 per cent effective, as some defects (under/over-sized beads) still end up within the placement jig. This requires a manual inspection to resolve defective wells.

The project with AMRC Cymru explored the feasibility of introducing automation into the process, using vision systems to inspect the quality of the bead tray, reduce errors and meet future demand by increasing throughput. 

Innovation: 

AMRC Cymru has developed two physical demonstrators, each utilising a different technology, to perform presence detection on racks of vials. By identifying which vials contained no beads, one bead, or multiple beads, workers would then be able to manually correct these issues.

The first demonstrator was a stationary camera connected to a Raspberry Pi, which utilised a deep learning machine vision model to scan multiple vials. It then classified whether a vial contained no beads, one bead, or multiple beads.

The main advantage of this demonstrator was that the whole system cost less than £500, however the deep learning model must be trained on each different type of bead and coloured vials.

The second demonstrator was a confocal displacement sensor, capable of measuring distances to micron level accuracy, mounted to the carriage of a 3D printer. By moving the carriage over vials, engineers were able to measure the internal dimensions of vials, and determine not only whether a bead was present, but also the size of the bead. 

This was paired with custom software which allows users to start automated inspection jobs and then view the results on a Human-Machine Interface (HMI). The advantage of this demonstrator was that it could detect the presence of beads regardless of the vial colour but the confocal sensor was expensive. 

Result:

The research team at AMRC Cymru successfully developed two demonstrators and could detect the presence of vials with 100 per cent accuracy in a process time of between 12 and 46 seconds.

Through this project, the team has also been able to expand its knowledge in 3D printing, deep learning machine vision, and confocal displacement sensor technologies. AMRC Cymru also developed its knowledge when it came to HMI technologies and is now more confident in developing HMI demonstrators.

Impact:

The work undertaken on this project proves that an automated inspection process is capable of meeting Biofortuna’s needs and delivering on high accuracy and repeatability, alongside deepening AMRC Cymru’s expertise in automated solutions involving advanced sensors and novel methodologies.

Recently, Biofortuna has been working with a robotics consultancy to implement an automated system to stack and move the 96 well plates along its production line and, after seeing the demonstrators developed by AMRC Cymru, it has expressed interest in integrating the demonstrators within the automated system that is in development. 

These developments mean that Biofortuna can increase the amount of product produced while maintaining quality standards, allowing for higher throughput of product, and enabling the reallocation of money and personnel to focus on more impactful tasks.