WE.2.E || Mobilizing LCA Resources Through Digital Collaboration

Thore, Andreas; Carlsson, Raul

Artificial intelligence (AI) can be a practical tool to enhance the LCA community’s ability to meet society´s demand for more life cycle assessment (LCA) and footprint studies. The reason is that LCA requires intense amounts of intellectual resources that makes LCA costly and time consuming, yet still invaluable for decision making for sustainable transformation. The paper discusses which strategy to apply to train artificial neural networks (ANN) so that the AI can get LCA right. A simple assumption about how an LCA AI may be used is that the AI is presented with a complex component and that it replies with the component’s life cycle profile or by proposing an environmentally better alternative. How should an ANN be trained to achieve this, and which data would it then need to present such results? In general an ANN is trained by providing it with some different and inexact data, and the ANN will learn to recognize not the data itself but some of its general patterns, so that it learns to put any new such data in a correct predefined category. To make use of AI to produce LCA:s we therefore need to know which inexact data and which correct categories to use to train an LCA AI. The paper discusses challenges with finding data to train the ANN into making the correct categorization, as well as which data the AI need to calculate LCA results. Challenges concerning use of training data from databases with old, average and calculated data, or from particular or unverified LCA-studies will be discussed. It will also be discussed what is needed if a company wants to have real time LCA results, to keep in pace with the technological advancements and continuously follow or lead the rapid sustainability improvements throughout its supply chain. The paper concludes that there are many ways to improve the way LCA is made by using AI, but there will be challenges to find verifiable data to train the ANN to know what is a correct LCA.