AI Research for the Development of Smart Factories for Connected and Electric Vehicles (PERTE SEAT-PP108)

The project "AI Research for the Development of Smart Factories for Connected and Electric Vehicles (PERTE SEAT-PP108)" aimed to generate new methodologies based on artificial intelligence (AI) to increase the competitiveness and efficency of the smart factories of OEMs and Tiers that make up the value chain of the electric vehicle of the future. 

To achieve thi goal, the consortium formed by ZYLK.NET, Fundación Vicomtech, Fundación Azterlan, WIP Proyectos industriales, Kaptura Inteligence, Kapture and Gestamp Abrera carred out a coordinated set of research initiatives aimed at transforming industrial processes through advanced digital technologies. The join efforts made it possible to address different areas of AI and its industrial application, following a comprehensive approach that ranges from data generation to intelligent automation and process control. 

Among the specific objetives pursued are: 

- Apply Explainable Ariticial Intellgience (XAI) techniques to develop interpretable models that provide useful insights for industrial business, fostering trust and data-driven decision-making. 

-Train Models in a descentralized manner, using knowledge transfer methods an semi-supervised learning, in order to reduce reliance on labeled data and improve system adaptability. 

-Develop new prediction methods capable of integrating signals or information soruces with a high stochastic compenent-such as demand, prices or operating conditions-and complementing them with external data to increase model reliability. 

-Design reinforcement learning systems that allow for the flexible and adaptive optimization of production processes on the shop floor, dynamically adjusting to context or demand changes. 

- Generate intelligent solutions with maximum reliability and precision, based on new Deep Learning models and volumetric analysis algorithms, aimed to the real-time detection of entities and events in compley industrial environments. 

- Investigate new AI architectures for industrial safety environments, combining 2D and 3D data acqusition and processing with deep learning models and volumetric analysis techniques, with the goal of creating safer and more automated work environments. 

- Develop a novel porous surface concept, pioneering intelligent vacuum management in high-pressure aluminum die-casting molds, allowing high-quality and weldable metallic components to be obtained. 

- Increase process capability toward the "zero defects" paradigm, especially in subsquent welding operations, through digitization and advanced monitoring of production parameters. 

- Investigate Industry 4.0 systems for the optimization of component and assembly flow, incorporating advanced analytics and process automation within the supply and distribution chain. 

- Automate loading and unloading processes through a new data acquisition method integrated into a digital simulation system, enabling intelligent control and traceability of operations. 

The execution of the project successfully achieved the proposed objectives, generating new scientific and technological knowledge in the field of artificial intelligence aplied to the manufacturing industry. Progress was made from a technology readiness level of TRL3 to TRL4, validating the solutions in laboratory environments and demonstrating their technical feasibility and potential for transfer to electric vehicle production lines. 

The project "AI research for the development of smart factories for connected and electric vehicles (PERTE SEAT-PP108)" was structured into five work packages grouping the R&D activities of the six participating partners, spanning from knowledge generation in artiificial intelligence to its validation in real industrial environments: 

1. Distribuited AI and Model Explanibility (WP1) 

2. Prediction and Flexible Process Optimization (WP2) 

3. Detection systems and Industrial safety based on deep learning (WP3) 

4. Digitization of Manufacturing processes and Vacuum control in Aluminum injection (WP4) 

5. Integrated Validation and Proof of concept (WP5) 

The project fulfilled all the scientific and technical objectives defined during the application phase, confirming the technical feasibility of the developed solutions and their potential for industrial scaling. 

The technologies for explainable IA, distributed learning, reinforcement optimization, vacuum digitization, and predictive contrl of industrial processes were validated in a laboratory environment (TRL4), meeting the planned milestones without significant deviations from the initial schedule. 

The consortium has consolidated a solid technology foundation that will enable advancement toward experimental development phases aimed at deploying smart, sustainable and competitive factories in the connected and electric vehicle value chain.