In the SOFIA project, Deep Blue and TXT are working together to develop an AI-based virtual environment for training industrial technicians. The application case: SIAD compressor maintenance technicians.
The European manufacturing industry is undergoing a period of profound transformation, driven by digitalisation and the growing incorporation of advanced technologies into production and organisational processes. According to the European Commission’s Digital Europe Programme, adopting digital technologies is one of the most effective ways to strengthen the competitiveness and resilience of the European industrial system. However, this transition has highlighted the growing need for skills, particularly in enabling technologies such as artificial intelligence, virtual reality, data analytics and intelligent automation.
In this context, vocational training plays a central role as a tool for professional development and a catalyst for innovation. According to Unioncamere’s 2024 data on digital skills, the mismatch between supply and demand is one of the most critical issues in the Italian manufacturing sector, with 60.1% of companies reporting difficulties in finding suitable professionals. The problem is equally evident internationally: according to the OECD, the manufacturing sector has the most widespread skills gaps, with 41% of companies reporting shortages compared to an average of 32% in sectors such as communications and finance. This reflects the rapid pace of technological change and the heavy dependence of the industrial sector on highly specialised technical skills.
Hence, the a need to experiment with new training models capable of transmitting knowledge in a more effective, immersive, and standardised manner. Technologies like artificial intelligence and virtual reality appear to be particularly promising solutions for this purpose.
SOFIA: A virtual environment for intelligent training
SOFIA is a project funded by the National Recovery and Resilience Plan (PNRR) that aims to develop a new generation of digital training tools in the manufacturing sector. Coordinated by the Italian group TXT, the project involves a consortium of public and private stakeholders—including CNH Industrial, SIAD Macchine Impianti, the Polytechnic University of Milan, and Deep Blue, a leading Italian SME for research and innovation projects carried out under the Horizon programs—with the goal of improving worker training through the integration of artificial intelligence and immersive virtual reality environments.
The core of the project is the evolution of the WEAVR digital platform, previously developed by TXT, which is enhanced with intelligent modules capable of observing operator behaviour, learning from real-world situations, and proposing optimised operational solutions. The artificial intelligence implemented is “explainable,” meaning it can make the suggestions provided to the worker understandable and verifiable, thus fostering trust and collaboration between humans and machines.
At the same time, SOFIA introduces immersive virtual reality interfaces, allowing operators to train in simulated environments that are faithful to real production, allowing them to safely repeat complex procedures. These simulations offer a more engaging and effective training experience, with the ability to monitor progress in real time.
A concrete case: the training of SIAD technicians
The SOFIA project finds one of its most significant applications within SIAD Macchine Impianti, an Italian company based in Bergamo, specialising in the design and production of compressors for industrial use. The goal is to train the technical staff responsible for the maintenance of these systems through an interactive virtual environment, enhanced by artificial intelligence.
“The developed system allows us to simulate maintenance operations in immersive environments, faithfully reproducing real work situations, including critical conditions or possible malfunctions – explains Andrea Capaccioli, Senior Consultant and Human-Centred Design expert at Deep Blue – Technicians can thus learn and refine their skills in a safe space, where it is possible to make mistakes, repeat and experiment without operational consequences”.
A distinctive element of the project is the integration of artificial intelligence into the training process: the system observes and analyses the operator’s actions during the procedure in a virtual environment, provides personalised feedback, and suggests corrective strategies. This makes learning more targeted and flexible, adapting to the needs of the individual worker.
Deep Blue’s contribution: designing interactions and training content
Within the SOFIA project, Deep Blue plays a central role in designing the learning experience and the interaction dynamics with the digital assistant, modelling AI behaviours to support different training approaches, from teacher to tutor or coach. Specifically, the company is responsible for the design of digital interfaces, the design of interactions between operators and AI, and the overall user experience of the system, combining AI and virtual reality. The goal is to design learning environments enriched by AI technologies for the training and continuing training of technicians and industrial operators.
The benefits for supply chain companies are numerous. First, they accelerate onboarding processes: new hires can acquire the necessary skills in a shorter timeframe thanks to practical, immersive and safe training. Furthermore, these solutions offer the opportunity to maintain the practice of less commonly used but crucial skills in specific contexts or extraordinary situations, thus contributing to staff operational readiness even in the long term.
An additional benefit is the ability to make tacit skills, often implicit in the experience of individual technicians, explicit and shareable. By identifying this knowledge and designing digital learning assistants, this knowledge can be made tangible and more effectively transmitted to the entire workforce.
Significant benefits are also seen in the learning and transfer of procedures, which minimises variability in the execution of operations and ensures greater quality and consistency in processes. This is accompanied by a reduction in human errors, thanks to a better understanding and familiarity with the tasks to be performed.
Finally, a distinctive element of Deep Blue’s contribution to the project is the development and application of a Human-AI Teaming model specifically designed for the training context. “This model allows us to precisely define the most effective ways for workers and AI to interact, articulating the user experience across multiple levels: from the graphical interface to interaction logic, to the most advanced forms of collaboration between humans and intelligent systems,” explains Capaccioli.
An approach that aims not only at training effectiveness, but also at building a relationship of trust and transparency with AI technologies.
