Human digital twins to realise production systems where human and machines complement their capacities to achieve better performance
Despite the increasing level of automation, workers’ characteristics, skills, behaviours and psychophysical conditions have a relevant impact on the performance and operations of manufacturing production systems. However, as of today, all these elements are almost neglected in the digital representation of the factory, not allowing to get the most from human and machines interaction.
Humans and automation systems must complement their capacities in order to achieve improved manufacturing performances, thus their historical data, status and evolution must be available for analysis and optimization. To achieve such a goal, it becomes imperative to create digital representations not only of production systems, but also of workers, considering context data (e.g. assigned job, current workplace, current shift, training program), quasi-static data (e.g. specific worker needs, skills, age) and real-time sensor data (e.g. accelerations, Heart Rate, Galvanic Skin Response, Temperature). All these data can be used to feed 1) monitoring models and algorithms focusing on specific features and attributes (e.g. detecting workers’ fatigue, estimating mental stress), 2) behavioural modules elaborating the workers’ current status to make predictions and simulate its evolution over time and 3) decision modules identify decisions in order to intervene in the digital and/or in the physical world.
Today, such human characterisation and ontology are only partial and not combined into the factory digital twin. The characterisation of the workers in terms of knowledge, skills, personal needs, intellectual and sensorial capacities and interactions with the factory entities is a cornerstone to actually develop an efficient human-factory relation. The true challenges lie in the fact that 1) humans need far advanced models to support their behavioural paths toward factory enhanced productivity and safety and 2) that these models need a complex understanding of the workers in terms of their features and behaviours.
In STAR, a Human Digital Twin is proposed in order to address these challenges. The STAR’s Human Digital Twin can be considered as a single source of truth of workers-related data. It has been conceived to provide a centralized access point to exploit an extensive set of workers’ related data, creating a digital representation of the workers, seamlessly integrated with digital factory, giving the possibility to AI-based modules to compute complex features, feeding and enriching the HDT itself, or to make better decisions, dynamically adapting automation systems behaviour targeting both production performances and workers’ safety and wellbeing.
 Bettoni, A., Montini, E., Righi, M., Villani, V., Tsvetanov, R., Borgia, S., ... & Carpanzano, E. (2020). Mutualistic and adaptive human-machine collaboration based on machine learning in an injection moulding manufacturing line. Procedia CIRP, 93, 395-400.