IMPROVEBridging Predictive Maintenance and
Machine Learning for Enhanced Performance
and SustainabilityReference Spoke Thematic AreaPartner CompaniesProject DurationFunding granted

Grant awardedMachine Learning (ML) models
and technological solutions
supporting predictive maintenance in
industrial applications and energy systems.DLVSystem SRL

Cal-Tek SRL12 months381.666,40 €

353.631,52 €→→→→
360° Predictive MaintenanceProcess Data→Predictive Maintenance AlgorithmsPredict machine downtimeMaintain high plant efficiencyMaintain high processed
product quality
The use of Artificial Intelligence
allows going beyond predictive maintenanceProduction process monitoring platforms + Machine Learning Techniques → Analysis ToolsBenefitsUser assistance in optimal
management of machine downtime

Proposing solutions
more suitable for the identified problemVerify in real-time if operators
are configuring the machine consistently
with the active production typeCreate "small intelligences" capable of
autonomously managing data control activities
and coordinating with each other.
What does the future hold?Representing the expertise
of sector experts within computersSoftware will be able to identify
causal links between machinery statesOptimize the multitude of parameters of interconnected
machinery, finding globally "optimal" configurationsInstead of "local" optimums obtained
by optimizing individual components of a production line→
The global market challenge requires
working with maximum efficiency
in every detail of workProduction processes have a complexity
that transcends human capabilityThe level of sensor technology,
the ability to process large amounts of collected data,
and to represent the information "distilled" from themjustifies the use
of Artificial Intelligence
techniquesnot to replace humans
but to assist them,
so that decisions are made
"objectively" and quickly→→→

IMPROVE Project: Leveraging Machine Learning for Predictive Maintenance and Sustainability in Manufacturing

CUP: D43C22003120001

COR DLVSYSTEM: 22675283

Start date: 09/2024 – End date: 08/2025

The IMPROVE project focuses on advancing predictive maintenance and promoting sustainability within the manufacturing industry by utilizing cutting-edge Machine Learning (ML) models and innovative technological solutions. Recognizing that modern manufacturing depends on complex machinery requiring diligent maintenance for peak performance, IMPROVE leverages ML algorithms to enable companies to shift towards proactive maintenance strategies, thereby enhancing overall operational management.

At its core, IMPROVE aims to implement a sophisticated predictive system designed to analyze data directly from Programmable Logic Controllers (PLCs) on the factory floor. This is achieved through a robust, interoperable architecture built on APIs and middleware, significantly enhanced by integration with the Fiware platform. Machine Learning algorithms are key components, driving the proactive approach to managing operations.

A major challenge addressed by IMPROVE is handling diverse (heterogeneous) data sources; the project tackles this by employing dynamic data models capable of real-time updates, boosting precision and adaptability. The use of middleware is crucial for ensuring interoperability between different systems and uniquely supports federated learning. This allows ML models to be trained effectively even with sensitive data that remains securely stored locally. Ultimately, this approach facilitates the integration of legacy and new systems into a single, unified platform, offering centralized data management and visualization capabilities.

The primary objectives of the IMPROVE project are multi-faceted: preventing equipment failures and degradation, optimizing operational reliability, making efficient use of resources, and significantly reducing the company’s carbon footprint. Tangible results expected include the ability to accurately predict machine downtime, maintain high levels of plant and equipment efficiency, and ensure the quality of manufactured products remains consistently high.

This 12-month initiative represents a collaborative effort between CAL-TEK, DLVSystem, and the University of Calabria. It aims to significantly advance the technology readiness level from TRL 1 to TRL 5. The project will culminate in the alpha release of the fully integrated platform within a production environment, with its functionality and progress validated through engagement with clients, internal staff, and other key stakeholders.