As governments around the world get more involved in fundamental technology legislation, some unplanned alignments are taking place. One of them is legislation intended to promote recycling and sustainable design meeting up with the pan-industry fashion for digital twins.
The past decade has seen even organisations like the World Economic Forum pushing the concept of the digital twin for everything from phone designers to entire cities. Though it only seems a small step from the model-based simulation electronics engineers have been doing for years, from SPICE through to Matlab’s Simulink, the twin goes somewhat further. The twin is meant to support the design from way past the point the team signs it off for manufacturing.
In principle, if you take data from the field and not just benchtop experiments you can make modifications to the model over time to iron out sources of early failure and inefficiency in successor designs. And you can tell customers who provide you with their own data what kind of maintenance their systems need. This is where the close companion to the digital twin comes in, and the link to some of the incoming legislation.
Digital threads capture the changing state of products and potentially the provenance of all the components they have ever contained. Being able to capture much of this data in the field has led to some industries changing their approach to business: renting out machinery and the support that goes with it. The most famous is the decision of Rolls-Royce to retain ownership of jet engines and use the in-flight instrumentation to schedule maintenance.
For jet engines and similar high-criticality equipment, the investment in the infrastructure needed to support a digital thread is easier to justify than in lower-cost products. Government-level changes may shift the economics.
Legislation to promote recycling and sustainability may have a far more widespread effect on the takeup of digital-thread techniques simply because it becomes worth implementing digital-thread systems to support it. Enforcement will phase in across different markets during the second half of this decade following an agreement in principle by EU member state on the Ecodesign for Sustainable Products Regulation (ESPR) at the end of last year.
The centrepiece of the legislation will be the digital product passport (DPP). Though the details are not yet clear on exactly how the EU will demand implementation, the DPP splits into two main components. The part consumers will usually see is an identifier stuck to the product. This could be as simple as a label with a QR code. Or it might be data stored in an RFID tag.
Rather than contain the passport data, the identifiers will simply link to a public record for the product: what it contains and what should happen to different parts at end of life. The data may also add provenance for the components to aid in carbon accounting.
Battery packs for electric vehicles are likely to be the first products to need a DPP in the EU, possibly as early as 2026. Manufacturers might not be forced to comply before the start of the next decade. But the EU has said it sees electronic and computer equipment as important targets for DPPs. The target products share a common issue: the sheer complexity of recycling and reusing what is inside them coupled with a reliance on minerals, such as lithium, that are likely to face regular scarcities in the pursuit of net-zero. The big exception will probably be medical devices.
Digital-thread concepts
Researchers on the Horizon 2020 CircThread project have argued that embracing DPPs should help manufacturers adopt digital-thread concepts. They can employ the information picked up when a product is resold or recycled on aspects of its use, such as how long it lasted before failing, even if they do not plan to collect data during the product’s lifetime.
Conversely, information from the DPP and its companion thread may help to prolong the useful life of products or subsystems by providing information on how to reprogram the devices safely.
Though manufacturers could just align each of the serial numbers stored in DPPs with the overall product design, obsolescence planning may encourage manufacturers to define digital threads for each unit. The internal details of individual systems and their provenance could differ from each other substantially and this data may prove essential to delivering lifetime software updates. Consumer-product vendors already have to deal with mid-production changes as part availability changes, which without a record of the changes demands over-the-air updates contain multiple images for the different device drivers.
A major issue for electronics suppliers will be how to aggregate the data needed to support a digital thread that can deliver comprehensive information on functions, recycling and provenance. One standard that could improve the situation is Jedec’s JEP30, which first appeared back in 2018. The standard defines an XML format for capturing product details and build what amounts to a machine-readable datasheet. EDA vendors such as Siemens have suggested putting links to online simulations of parts into the JEP30 files. For security, the standard supports the ability for manufacturers to sign each element digitally and so flag up attempts to forge or alter the information.
There is an inevitable inertia to converting the mass of information in PDF datasheets and internal databases into a JEP30-compatible form. PCB-design specialist Flux sees machine learning as one way to move from these unstructured documents to JEP30 form and from there into digital-thread databases.
Recent developments in the way manufacturers approach the design of complex ICs may put more momentum behind the standard, at least in one potentially large and well-funded community: data-centre systems. In trying to find a way of corralling the many pieces of information needed to assemble chiplets into system-in-package (SiP) modules that will go into high-performance servers, the Open Compute Project (OCP) looked for existing standards members could use.
Jawad Nasrullah, Palo Alto Electron CEO, said at the OCP Summit last year, one reason behind developing machine-readable standards for chiplets was to make it easier to build generative algorithms that organise them into package designs automatically. In principle, EDA tools could ingest the data and then create layouts to minimise thermal hotspots and signal latency across the package for a variety of different configurations that are then assembled to order at the fab.
That machine-readable standard turned out to be JEP30, with some additions. At the end of May, Jedec released the JEP Part Model Guidelines for chiplets. These guidelines incorporate an XML format developed by the OCP for handling some of the information not already in JEP30.
Though JEP30 will mostly support design and early manufacture, doubts about the long-term reliability of chiplet-based modules, particularly the largest designs that consume a large proportion of a silicon wafer, may well tie into digital-thread databases. One major concern lies in temperature-induced warping that could cause the devices to separate from the substrate. A combination of JEP30 materials and manufacturing information coupled with data on temperature and other conditions captured during operation could prove vital in the design of successive generations.
When the time comes to retire the processor complexes, which given the pace of AI development could be just a few years, the data captured in JEP30 will probably prove useful to upcyclers looking to use the modules in other high-performance computing applications.
Though the digital thread remains in its infancy, the coming together of these different strands may prove instrumental in the rate at which the technique develops.
