Designing flexible automation
Most medical devices are assembled and tested using automated equipment, but it’s not unusual to need to specify and build the assembly equipment while the design of the device itself is still evolving. The problem for system developers is providing the necessary flexibility within the automation to accommodate these adaptations, while at the same time delivering the required control of process when the final medical device moves into production.
This dilemma is particularly relevant for medical device manufacturing because the possible impact on human health of a faulty assembly means that almost every step in assembly and testing must be shown to be operating correctly within a well-defined range of process parameters as part of a rigorous process of validation.
Let’s use the example of a probe used in keyhole surgery, which must be tested thoroughly at the end of manufacture to prove correct assembly and operation, to highlight five possible issues that can arise in development of a special purpose machine. And to illustrate how the system developer and device manufacturer can work together to mitigate them.
1 Pre-production samples are never truly representative
Samples of product used for design and development of the automation may differ in unpredictable ways from the final version, even when the product design and specification has not changed, and these differences will affect the operation of the automation equipment.
For example, the end of line test sequence for the surgical probe incorporates an electrical test function which worked as expected throughout pre-delivery commissioning and testing, but started to yield unexpected failures once the machine was delivered to the manufacturing site. It was eventually discovered that the spurious failures were arising from the inductance of the cable that is part of the device under test. The samples provided for factory development and testing were identical to those now showing the problem but their cables had been looped differently. Once recognised, this particular issue was easily resolved by a simple change to the configurable test sequence, but it could have been diagnosed far more quickly if the focus had been on the product as well as the machine.
A further lesson from this experience is to keep samples sets used within previous testing so they can be referred to again if necessary.
Dr Peter Woods is head of medical device manufacturing at GB Innomech but has helped many global businesses to avoid the unnecessary costs and delays due to manufacturing issues in bringing new products to market.
2 Automated measurements may differ from benchtop tests
An example is where a flow test, carried out as one of a sequence of pneumatic tests, gives lower values than the same test carried out “on the bench”. This difference arises because of the flow impedance of the much longer pipework runs combined with solenoid valves that are not included in a benchtop test system setup.
This is unlikely to be a major issue, but may require some extra validation work to reconfirm the relationship between the machine measurements and those obtained by independent testing. However, it will lead to shifting the acceptable range of a measurement in the automated test system, which may not be as simple to deal with as it first appears.
The lesson here is to anticipate that pass bands may need to be revised on the basis of empirical tests when the final product and machine are available.
End of line testing equipment – for example for multi-dose injector pens as shown here – needs to be designed with flexibility built in so that its pass criteria can be altered to accommodate product design changes.
3 Late changes to product designs may impact the automation
The surgical probe example features a slider assembly that is ergonomically-designed to be operated by the surgeon with minimum and maximum acceptable forces. This range may be quite broad but the device itself will involve a much narrower range, centred on the particular properties of the slider as implemented in that particular design. To confirm correct assembly, the measurement system will need to be configured to apply a narrow pass criterion corresponding to the final design.
The pass criteria themselves clearly need to be configurable and production staff should only be able to vary the pass band within a validated range. However, this validated range may itself need to change if the device design changes and so this should also be configurable by appropriate authority. But for maintenance (to diagnose faults) and software validation (to elicit specific conditions) it may be desirable to set a pass/fail condition that is outside the validated range.
This dilemma can be resolved by avoiding the temptation to specify absolute values for limits (and limits on limits) in design specifications and definitely not to hard code these values but instead to define a different level of access to these parameters to different classes of users. This creates additional work for the system developer but the payback is swift: the finally agreed values for allowed ranges can then be put in place as a matter of simple configuration, rather than a series of updates to code and document.
Maintenance functions should be designed in as a permanent feature of the final delivered system so that any future problems with individual process steps can be isolated, accurately diagnosed and resolved.
4 Product design changes to facilitate automation can pay off
The early design of the product cannot anticipate the detail of how components are handled in assembly and testing, and so making minor changes could significantly simplify the automation and improve reliability. The cost of introducing a change in design is a genuine barrier, but the design may evolve for other reasons. And in retrospect the system designer will often think: “If we had known it was going to change anyway, we would have insisted on our requested change”.
In our example, a matching nest carries the assembly through the production machine, but the symmetry of the design means that the part can sit at an angle, potentially causing alignment issues in assembly. A change to the moulding, combined with a corresponding detail in the nest, would completely eliminate this risk but changes to mould tooling are expensive and take time.
Since more and more prototyping is now being carried out by 3D printing, there is scope to defer the commitment to the final tooling and more opportunity to incorporate features to facilitate automated handling. This is something both system designers and device producers should bear in mind.
Engineers should avoid defining the detailed manufacturing and test processes too early, particularly when the product itself is not final. In the example shown above, an extra process step to check UV glue spots which was not initially specified was found to be essential during system development.
5 Design the machine with validation in mind from the start
Designing a system solely to meet the User’s Requirements Specification (URS) will result in a machine that is difficult to validate.
The system designer should anticipate validation throughout the design phase so that the assembly machine not only meets the explicit user requirements but also provides functionality to control and monitor operation of machine elements, simplifying maintenance procedures and avoiding where possible the need for painstaking validation of software by exhaustive testing.
In our example, one aspect of validation would be to verify that the value of slider force recorded by the machine truly corresponds to the actual force applied. The most straightforward implementation might involve a calibrated load cell outputting an analog signal that is digitised and captured by the machine control system. But this involves two independent calibrations (the load cell and the analog/digital conversion) and where the conversion of sensor output to displayed and recorded force is hidden within the software.
The first important feature needed within the machine is a maintenance function that will allow the force measurement to be carried out in isolation. The details of maintenance functions will not be defined in a URS because they are specific to the assembly machine, but providing substantive maintenance functionality will support not only machine commissioning and servicing but also validation and functional testing.
In addition, using a sensor that produces a serial output already converted into the relevant units (Newtons in this case) greatly simplifies calibration and validation. There is now only one subsystem to be calibrated, which can be removed from the machine for that purpose when necessary, and the values it produces can be seen and monitored to easily prove exact correspondence with the values recorded by the machine.
In summary, failing to anticipate the possibility of changes in the product design can lead to costly reworking of design documentation and implementation. By contrast, by anticipating this possibility and by working closely with the manufacturer, any unexpected changes can be accommodated much more easily and adaptations that simplify automation can also be considered.
Similarly, focussing only on the need to have the process “locked down” in production can lead to difficulties when attempting to prove the operation of machine elements for validation or in later fault diagnosis. But if system developers can keep these maintenance and testing requirements in mind from the start of design then they can obtain a payback in flexibility of operation whilst still meeting all the user requirements.
About Dr Peter Woods
Dr Peter Woods is head of medical device manufacturing at GB Innomech and leads a multi-disciplinary team to design and build automation equipment for developers and manufacturers of products such as drug delivery pens, contact lenses, diagnostic kits and surgical instrumentation.
He is an acknowledged expert in pharmaceutical automation and particularly skilled in system analysis and performance validation to help companies avoid unnecessary costs and delays due to manufacturing issues that can occur when bringing new products to market. He has helped many global businesses, not just pharmaceutical companies to use automation to reduce costs, while driving up quality and productivity. He has a PhD from University of Manchester and in 2008 was named as a member of one of the four finalist teams shortlisted for The MacRobert Award, from The Royal Academy of Engineering for his pioneering work on the Polar system for UK Biobank.
About GB Innomech
GB Innomech (Innomech) specialises in automating highly complex and labour-intensive manufacturing processes to maximise outputs, improve product quality and boost business performance. The company works with major international manufacturers in sectors such as pharmaceuticals, medical devices and environmental, as well as earlier-stage businesses looking to bring breakthrough technologies or products to market.
Innomech has a growing market reputation for solving the toughest of manufacturing problems by the early identification and management of risk, often cross-fertilising technologies and techniques from a range of industry sectors. All projects from initial feasibility studies through to building production-scale machines are conducted to high specification pharmaceutical industry standards and are designed to comply with GAMP5, FDA and other international standards.
The company was founded in 1990, is based at The Innovation Centre in Witchford, north of Cambridge and was awarded The Queen’s Award for Enterprise 2009 to recognise its sustained growth in international markets.
For additional information about GB Innomech please visit or contact:
• All other enquiries to Tim Mead at Innomech on +44 (0)1353 667394