EXPlainable and Learning production & logistics by Artificial Intelligence

The project conducts research and development of a new generation of interactive and innovative fusion of virtual production modeling methods and machine learning algorithms for decision-making support and increasing knowledge and competence within the production systems lifecycle. EXPLAIN will target cases on production planning and control with humans-in-the-loop, wherein complex multi-criteria decisions are to be made, including energy and resource efficiency.

The EXPLAIN project also brings into a paradigm that emphasizes the human-machine co-learning through transferability of preferences/values and knowledge between human and machine within a multi-objective (productivity and sustainability) optimization context and hence will provide a unique, long-term contribution to “knowledge-driven industry” in Sweden. Fully in-line with other worldwide Learning Factories efforts, the “human-machine symbiosis” framework proposed by EXPLAIN can in the long-term increase the sustainability and competitiveness of the Swedish manufacturing industry.

Background

Transforming information and data into knowledge and decision support using virtual production technologies systems represents one of the major industrial challenges today. On the one hand, the use of virtual production technologies in the industry is increasing with the rapid development of information and communication technologies (ICT). Their applications have been able to shorten both development time and cost because they allow experimentation with different designs or changes, without using any real process or equipment. On the other hand, optimal and sustainable decision-making by managers will not automatically come from the abundant use of virtual production models and the data ubiquity alone.

First, virtual models can only provide performance evaluations of different design alternatives and/or courses of actions so they are not real optimization tools. It is when virtual models are connected to Artificial Intelligence (AI) based optimization algorithms that optimal solutions can be sought automatically. Second, sustainable decisions cannot be made if sustainability factors, like energy and resource efficiency, are not considered in the virtual production modeling process. Third, building complex virtual models requires the right level of competence and skills, which can be a time-consuming and costly process if done manually. It would be even more complex when the virtual models have to take into account the sustainability factors. It is also highly desirable if the knowledge and experiences gained when building and using virtual production technologies can be easily searched, retrieved, and reused within the company.

Project idea

Therefore, unlike many other research efforts within virtual production, EXPLAIN aims at delivering three unique innovations: (1) automatic virtual model generation with real-time connections to data sources and modeling sub- modules that consider energy and resource efficiency factors via process mining and self- reconfigurable modeling methods to further reduce the hurdle and lead-time of using virtual production technologies in the industry; (2) explainable and trustable AI algorithms and user interfaces for decision-makers, not only for informed, optimal, and confident decisions but also for increasing their knowledge and competence; (3) knowledge management using knowledge graphs to link virtual production artifacts to enable them to be searchable, retrievable, and efficiently reusable by human users.

Partners

EXPLAIN brings two of the top-ranked universities in Sweden, Uppsala University and KTH, along with four highly competitive Swedish manufacturers from very different sectors: Scania, Hitachi-ABB, AstraZeneca, and SECO Tools, together with MainlyAI as a startup SME in AI, as well as research institute RISE, to form a highly diversified consortium. The inherent diversity of the industrial partners from four sectors, together with the roles of MainlyAI and RISE in implementing and disseminating the delivered innovations, will facilitate them to be immediately applicable to a broad range of production plants in Sweden.

Project leader Amos Ng, Uppsala University.