This thesis contributes to the debate of digital twins for battery manufacturing by systemically classifying and analysing cutting-edge models of the literature, as well as developing an innovative discrete events model of a battery production line. It holds significant relevance because it addresses the urgent global need to expand energy storage capacity and consequently to increase and improve battery production. Our goal is to bridge two existing gaps of the literature about digital twins for battery production: the lack of systematic studies on digital twin technology for battery production and the scarcity of tools for modelling the complex nature of battery manufacturing, which involves continuous, discrete, and batch operations. So far, the current literature demonstrates an absence of validated production models and a lack of clarity and transparency in the assumptions and parameters that drive these simulations. This thesis presents and describe the modelling techniques of the state-of-the-art, including physics-based models, machine learning models, production models, multilevel models, and digital twins, as well as how they are utilized to forecast battery features and production performances. Our validated discrete event model not only captures the complexity of battery manufacturing, but it also is based on reasonable explicit assumptions and well-defined parameters. The proposed model is used to conduct numerical studies on throughput. In the first analysis, our goal is to determine if the factors of processing time, time to failure, and time to repair for both the cutting and stacking machines, as well as the buffer capacity between these production stages, are relevant in increasing productivity. This is performed using a Design of Experiments approach based on Latin Hypercube. The output analysis is conducted using Generalized Linear Model. This analysis give insight into the role of each of the seven factors in determining the throughput. In the second study, through the application of a second Latin Hypercube, a set of values of the relevant parameters that ensures the throughput exceeds 0.48 cells per minute is identified. Therefore, this thesis contributes to the academic discussion of digital twins for battery manufacturing while also provides practical ideas and approaches that may benefit the battery industry.
Questa tesi contribuisce al dibattito sui digital twins per la produzione di batterie classificando sistematicamente e analizzando i modelli all'avanguardia presenti nella letteratura, nonché sviluppando un modello innovativo ad eventi discreti di una linea di produzione di batterie. Ha una rilevanza significativa perché affronta l'urgente necessità globale di espandere la capacità di immagazzinamento energetico e, di conseguenza, di aumentare e migliorare la produzione di batterie. Il nostro obiettivo è colmare due lacune esistenti nella letteratura sui digital twins per la produzione di batterie: la mancanza di studi sistematici sul tale tecnologia e la scarsità di strumenti per modellare la complessa natura della produzione, che include operazioni continue, discrete e in lotti. Finora, lo stato dell’arte dimostra una mancanza di modelli di produzione validati e una carenza di chiarezza e trasparenza nelle ipotesi e nei parametri che guidano tali simulazioni. Questa tesi presenta e descrive i modelli della letteratura ovvero modelli basati sulla fisica, modelli machine learning, modelli di produzione, modelli multilivello e digital twins, nonché come sono utilizzati per prevedere le caratteristiche delle batterie e le prestazioni di produzione. Il nostro validato modello ad eventi discreti non solo cattura la complessità della produzione di batterie, ma si basa anche su ipotesi esplicite ragionevoli e su parametri ben definiti. Il modello proposto è utilizzato per condurre studi numerici sul throughput. Nella prima analisi, l’obiettivo è determinare se i fattori tempo di lavorazione, tempo al guasto e tempo di riparazione per entrambe le macchine di taglio e impilamento, così come la capacità del buffer tra le due, sono rilevanti nell'aumentare la produttività. Ciò viene eseguito utilizzando un approccio di Design of Experiments basato su Latin Hypercube. L'analisi dei risultati è condotta utilizzando Generalized Linear Model. Questa analisi fornisce intuizioni sul ruolo di ciascuno dei sette fattori nel determinare il throughput. Nel secondo studio, attraverso l'impiego di una nuova analisi basata su Latin Hypercube, sono stati identificati un insieme di valori per i parametri rilevanti che assicurano che il throughput ecceda la soglia di 0.48 celle al minuto. Questo lavoro contribuisce quindi alla discussione accademica sui digital twins per la produzione di batterie ed offre spunti pratici per l'industria.
Digital twins for battery manufacturing
Gastaldi Cibola, Riccardo Maria
2022/2023
Abstract
This thesis contributes to the debate of digital twins for battery manufacturing by systemically classifying and analysing cutting-edge models of the literature, as well as developing an innovative discrete events model of a battery production line. It holds significant relevance because it addresses the urgent global need to expand energy storage capacity and consequently to increase and improve battery production. Our goal is to bridge two existing gaps of the literature about digital twins for battery production: the lack of systematic studies on digital twin technology for battery production and the scarcity of tools for modelling the complex nature of battery manufacturing, which involves continuous, discrete, and batch operations. So far, the current literature demonstrates an absence of validated production models and a lack of clarity and transparency in the assumptions and parameters that drive these simulations. This thesis presents and describe the modelling techniques of the state-of-the-art, including physics-based models, machine learning models, production models, multilevel models, and digital twins, as well as how they are utilized to forecast battery features and production performances. Our validated discrete event model not only captures the complexity of battery manufacturing, but it also is based on reasonable explicit assumptions and well-defined parameters. The proposed model is used to conduct numerical studies on throughput. In the first analysis, our goal is to determine if the factors of processing time, time to failure, and time to repair for both the cutting and stacking machines, as well as the buffer capacity between these production stages, are relevant in increasing productivity. This is performed using a Design of Experiments approach based on Latin Hypercube. The output analysis is conducted using Generalized Linear Model. This analysis give insight into the role of each of the seven factors in determining the throughput. In the second study, through the application of a second Latin Hypercube, a set of values of the relevant parameters that ensures the throughput exceeds 0.48 cells per minute is identified. Therefore, this thesis contributes to the academic discussion of digital twins for battery manufacturing while also provides practical ideas and approaches that may benefit the battery industry.File | Dimensione | Formato | |
---|---|---|---|
2024_04_Gastaldi.pdf
solo utenti autorizzati a partire dal 19/03/2027
Dimensione
3.58 MB
Formato
Adobe PDF
|
3.58 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/10589/217964