Getting ready for predictive lifecycle assessment models

We will one day be able to predict sustainability impact of design changes at the push of a button. We need to start preparing our data and IT now.

In previous articles, we discussed the importance of measuring the environmental impact of complex products – such as planes, trains and cars. The next big thing in sustainable engineering, we believe, is to move beyond manual data collection, and automate it via a model-of-models. This would allow design engineers to experiment in silico, making ‘virtual’ changes to design, materials, or supply chains, and immediately understanding how combinations of decisions change the overall lifecycle environmental impact.

The idea is to create live models of every aspect of your product’s lifetime environmental impact – from raw materials and manufacturing footprint, to in-use emissions, to end-of-life disposal or better yet re-use in a 2^nd life. Then, to build an overarching model that combines all of these models into one. This would dramatically improve their ability to make sustainable design choices.

Moving from life cycle analysis to life cycle modelling

This challenge is often underestimated. Many think data management software and access to SAP is all that will be needed. But in reality, significant is work needed on data and IT architecture, as well as supplier and customer engagements, before these sub-models – let alone the overarching model – can be reliably built, connected, and trusted.

Life Cycle Assessment (LCA) tools for reporting and even planning are advancing (we have a methodology for calculating the carbon impact of projects, for example). However, Capgemini Research Institute (CRI) research found that 45% of organizations are not using their emissions data for decision-making in any way, beyond mandatory reporting. No one we are aware of is successfully using autonomous tools that take such data and use it to support decisions by modelling their impact on complex systems.

However, progress is being made and best practice is starting to emerge from early experimentation.

Gather the right data in the first place

The first challenge is gathering all the data you need for the sub-models.

Your own emissions (Scope 1 & 2) can be captured by deploying electricity and gas meters, weighing fuel, tracking vehicles, and energy invoices. That, combined with data on fuel values on the local energy mix, can be used to build models that calculate emissions. This is not technically challenging, but deployment can be a sizeable project in a large organisation.

Emissions beyond your organisation – supply chains, product emissions, end-of-life (Scope 3) are trickier.

For in-use emissions, products such as cars and planes now have sensors which collect detailed usage data. That data can feed physics-based models to derive energy use and emission, which can be updated in near real-time. Products without such sensors, however, will need to rely on models which approximate their impact.

Suppliers, manufacturers, and disposal are harder to collect data on. Whilst some suppliers do their own LCAs, many do not collect even basic energy data, and there is little international standardisation.

Some may respond to encouragement, especially if you are one of their big customers. Workshops and guidance on what you need may help, as may paying to install sensors at their site, share product and material related information, or access to shared reporting software that feeds your own supplier models. Sticks may support carrots, such as audits and threats to switch to suppliers with better environmental data. Making PCF (Product carbon footprint) data reporting a condition for any new customer will help in future.

If all else fails, there are industry benchmarks to fall back on to calculate emissions of materials and parts, based on the materials and local energy mix where they were mined and processed.

A particular challenge is integrating new concepts. Creating values for different steel types is not too tricky, since there is lots of historical data. Understanding the impact of an untested biomaterial is harder. Evaluating the impact of a whole new technology, such as hydrogen, is a real challenge. There’s some chicken and egg; we need data to make the projection, but we want to project before we make the investment. The best middle ground is to make sensible projections based on scientific and engineering expertise, then gather data as the product evolves, which feeds directly back into the model to improve its predictive power.

Cleaning and clarifying your data

All the data coming into your models needs to be well-defined, consistent, and in machine-readable formats.

This starts with setting consistent policies for data collection and formats across your own organisation, and where possible communicating these to your value chain. An industry-led project in automotive, Catena-X, provides a good model for how data may be shared across the supply chain in future, and designing data capture and modelling that will integrate with this ecosystem is advisable.

That may set a path for the future. But a lot of legacy data – internal and external – has evolved in silos over the years, from engineering data, to excel spreadsheets, to PDFs of technical drawings. That will necessitate an exercise to find, clean, convert and tag data.

AI tools can – crawl IT systems, screen for the right data, and pull it from PDFs, Excel and so on, checking it, filling in gaps with industry standard figures, and pumping it out in a format that is Creating the right IT infrastructure for a model-or-modelsreadable by your model.

Creating the right IT infrastructure for a model-or-models

All of this data is spread out in different parts of the organisation, suppliers, and customers. But we want a single source of truth so that all data will be ‘Findable, Accessible, Interoperable, and Reusable’ or FAIR.

That means setting up a sustainability data hub – a master database in the cloud – where all relevant data is fed and validated.

It will also need software customisation to ensure all sources of data – whether sensor management platforms or CRM databases – are collecting data in the right format and updating the master database in real-time.

For sharing data across supply chains, or between manufacturers and customers, privacy and cross-border data-sharing rules also need to be considered. Blockchain-based databases offer good solutions to tracking parts and products securely as they move along the value chain. We already see companies like BMW using blockchain to track supply chains, and Siemens has a blockchain-based tool that lets suppliers share verified emissions data with customers, whilst keeping any underlying sensitive data confidential.

Conclusion: Aim high, create value along the way

All of this will be very bespoke to each organisation. It will mean working with data and software experts, as well as domain experts familiar with the materials and processes that the data represents.

This will not happen overnight, but the journey will also provide value. The best strategy is to have an eye on long-term value, whilst delivering more immediate returns. Improving data, building sub-models, and connecting up data streams will help life cycle assessments and small-scale modelling projects, as you gradually build to a systems-level model-of-models.

Ultimately – as companies test ideas in silico and see how they ripple through the supply chain and product life – they will become better able to make smarter, and often disruptive, decisions in sustainable design. No one is there yet, but this is the direction of travel.