n recent years, manufacturing has experienced a major transformation driven by digital technologies, advancing towards the concept of Industry 4.0. Among the innovations, Dig- ital Twin (DT) systems are increasingly recognized as valuable tools to improve production performance and to support operational decisions. A DT creates a virtual representation of a real manufacturing process and allows its behaviour to be monitored and analysed in real time. To ensure the usefulness of this technology, it is essential to keep the dig- ital model correctly aligned with the physical system. For Discrete Event Simulation (DES)-based twins, this alignment requires continuous monitoring, since conventional of- fline validation techniques are not sufficient in rapidly changing industrial environments. This thesis proposes a framework designed for the continuous validation and calibration of DES models in manufacturing. The method is based on comparing sequences of events and performance indicators between the real and simulated systems, diagnosing the poten- tial causes of discrepancies, and updating model parameters to maintain consistency. By combining these steps, the approach creates a feedback process able to promptly identify when the model diverges from reality and to adapt it accordingly. The methodology has been tested in two stages. First, controlled pilot studies with simplified systems were car- ried out to examine the capacity of the framework to detect and correct deviations. Then, the approach was implemented in a laboratory-scale production line involving multiple stations, where a proof-of-concept Digital Twin was developed. The results demonstrate the feasibility of the method and confirm that it can sustain the accuracy of DES-based Digital Twins in real time.
Negli ultimi anni, il settore manifatturiero ha vissuto una profonda trasformazione gui- data dalle tecnologie digitali, evolvendo verso il paradigma dell’Industria 4.0. Tra le innovazioni, i sistemi di Digital Twin (DT) sono sempre più considerati strumenti fonda- mentali per migliorare le prestazioni produttive e supportare i processi decisionali. Un DT consente di creare una rappresentazione virtuale di un processo produttivo reale e di analizzarne il comportamento in tempo reale. Per garantire l’efficacia di questa tecnologia, è indispensabile mantenere un corretto allineamento tra il sistema fisico e il modello digi- tale. Nel caso dei gemelli digitali basati su Simulazione ad Eventi Discreti (DES), questo allineamento richiede controlli continui, poiché le tradizionali tecniche di validazione of- fline non risultano adeguate in contesti industriali dinamici. La presente tesi propone un framework dedicato alla validazione e calibrazione continua dei modelli DES in ambito manifatturiero. L’approccio si basa sul confronto delle sequenze di eventi e degli indica- tori di prestazione tra sistema reale e sistema simulato, sull’identificazione delle possibili cause di discrepanza e sull’aggiornamento dei parametri del modello per mantenere la coerenza. Grazie a questa integrazione, il metodo realizza un ciclo di feedback capace di individuare tempestivamente le divergenze e di adattare il modello in maniera autonoma. La metodologia è stata sperimentata in due fasi. In primo luogo, sono stati condotti studi pilota su sistemi semplificati per valutare la capacità del framework di rilevare e correggere le deviazioni. Successivamente, l’approccio è stato applicato a una linea pro- duttiva di laboratorio composta da più stazioni, in cui è stato sviluppato un Digital Twin dimostrativo. I risultati hanno confermato la fattibilità del metodo e mostrato come esso possa garantire l’accuratezza dei gemelli digitali basati su DES in tempo reale.
Real-time validation and online input parameters calibration within a digital twin framework: an approach based on bootstrap particle filtering
Dhore, Abdirahman Said
2024/2025
Abstract
n recent years, manufacturing has experienced a major transformation driven by digital technologies, advancing towards the concept of Industry 4.0. Among the innovations, Dig- ital Twin (DT) systems are increasingly recognized as valuable tools to improve production performance and to support operational decisions. A DT creates a virtual representation of a real manufacturing process and allows its behaviour to be monitored and analysed in real time. To ensure the usefulness of this technology, it is essential to keep the dig- ital model correctly aligned with the physical system. For Discrete Event Simulation (DES)-based twins, this alignment requires continuous monitoring, since conventional of- fline validation techniques are not sufficient in rapidly changing industrial environments. This thesis proposes a framework designed for the continuous validation and calibration of DES models in manufacturing. The method is based on comparing sequences of events and performance indicators between the real and simulated systems, diagnosing the poten- tial causes of discrepancies, and updating model parameters to maintain consistency. By combining these steps, the approach creates a feedback process able to promptly identify when the model diverges from reality and to adapt it accordingly. The methodology has been tested in two stages. First, controlled pilot studies with simplified systems were car- ried out to examine the capacity of the framework to detect and correct deviations. Then, the approach was implemented in a laboratory-scale production line involving multiple stations, where a proof-of-concept Digital Twin was developed. The results demonstrate the feasibility of the method and confirm that it can sustain the accuracy of DES-based Digital Twins in real time.| File | Dimensione | Formato | |
|---|---|---|---|
|
2024_25_Dhore_Executive Summary_02.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
6 MB
Formato
Adobe PDF
|
6 MB | Adobe PDF | Visualizza/Apri |
|
2024_25_Dhore_Thesis_01.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
9.27 MB
Formato
Adobe PDF
|
9.27 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/243970