1. Context and Objectives of the Project

The STEELSEED project arises from the need to develop cross-cutting digital solutions based on explainable artificial intelligence (AI) for the improvement of energy efficiency, production quality and industrial maintenance in machining environments. The focus of the project is to advance the digitalization of the industry through a SaaS platform with monitoring, predictive analysis and decision support capabilities.

Main Objectives:

• Research distributed architectures and AI-enabled SaaS platforms to improve efficiency and quality in industrial processes.
• Develop advanced explainable AI techniques for fault prediction, process analysis and OEE/OLE improvement.
• Study non-destructive methods of automated quality control.
• Validate the solution in a simulated environment representative of part machining.

2. Project Partners

The consortium is composed of:

• FAYMM: Specialized in solutions for mobility, logistics and efficient resource management.
• SISTEM (CPS Group): Focused on intelligent transportation, defense and telecommunications.
• DATISION (Advanced Algorithms): Provider of explainable data analysis and AI solutions.
• GDE (Consultant): Coordination, support and supervision of the project.

3. Activities Carried Out — Milestone I

Activity 1: Definition and Scope

• State of the art of explainable AI in the industrial field (OEE/OLE and energy efficiency).
• Architectural design based on Federated Learning, structured in layers: acquisition, business logic and presentation.

Activity 2: Improving the OEE/OLE

• Development of hybrid neuro-symbolic algorithms for operations optimization.
• Non-destructive analysis of parts, characterization of critical parameters and preparation for data ingestion.
• Application of techniques such as clustering (DBSCAN, KMeans), Isolation Forest and SHAP for anomaly detection.
• Failure, performance and faulty part prediction models using algorithms such as Decision Tree, SVR and XGBoost.
• Forecast of OEE/OLE metrics using ARIMA, Prophet and AutoGluon with datasets up to 4 years old.

Activity 3: Improving Energy Consumption

• Design of the energy monitoring system at FAYMM, covering specific machines and electrical subsystems (grid, photovoltaic, air conditioning).

4. Milestone I Technical Results and Achievements

SYSTEM:

• Initial design of the federated learning architecture that will serve as the basis for the STEELSEED platform.
• Integration with the KEPSERVER platform through OPC-UA, FOCAS and MTConnect.
• Development of explainable AI algorithms for the calculation and prediction of OEE/OLE metrics.
• Definition of hardware requirements for energy monitoring.

FAYMM:

• Connectivity and IP configuration of ten machine tools for data capture.
• Identification of critical points in the machining process for analysis and modeling.
• Installation of sensors and devices for collecting electrical and energy data.

DATISION:

• Review of the state of the art in explainable AI and federated learning.
• Development of predictive algorithms on operational, maintenance and quality data.
• Evaluation of models with metrics such as MSE, MAE, false positive rate and correct prediction of faulty and stopped parts.

5. Conclusions

The STEELSEED project has successfully achieved the technical objectives set out in the first milestone. It has been achieved:

• Prototype a robust federated architecture for industrial environments.
• Integrate explainable AI technologies applied to operational and energy efficiency.
• Establish the necessary sensing and connectivity infrastructure in the real environment (FAYMM).
• Obtain functional predictive models that allow us to detect operational deviations and anticipate production incidents.

These advances consolidate the technical basis for deploying a scalable and transferable solution to other industries.

‍