Manufacturing companies have been operating in more and more dynamic contexts, and the overall goal is represented by the ability to act proactively and reactively to unpredicted and disruptive events, at all decision levels. Model-based methods allow what-if analysis, as well as the evaluation of situations that have never been observed in practice. In the literature, stochastic approximate analytical methods represent a class of models for performance evaluation of manufacturing systems providing useful intuitions about system behavior but often with restrictive assumptions. When continuous models are used for the performance evaluation of discrete manufacturing systems, they might introduce significant inaccuracies. This work presents a stochastic approximate continuous analytical method for evaluating the performance of asynchronous discrete manufacturing systems. The novelty consists in the integration of control mechanisms that allow the modeling of the peculiar features of discrete manufacturing systems in the continuous domain. The model is based on continuous-time mixed continuous- and discrete-state Markov Chain representation. A novel set of equations to evaluate the propagation of system-level phenomena is provided. Numerical validation shows good results for the model compared to discrete-event simulation. The applicability in an industrial context is demonstrated through the analysis of two real cases. The first case includes the evaluation of an Opportunistic Maintenance policy in a manufacturing line producing components for ready-to-assemble furniture. The second case analyzes different restart policies in a food production system characterized by mixed continuous-discrete production, with the goal of improving production quality.
Le aziende manifatturiere operano in contesti sempre più dinamici, e l'obiettivo generale è rappresentato dalla capacità di agire in modo proattivo e reattivo rispetto a eventi imprevedibili e dirompenti, a tutti i livelli decisionali. I metodi basati su modelli consentono analisi what-if e la valutazione di situazioni che non sono mai state osservate nella pratica. Nella letteratura, metodi analitici approssimativi stocastici rappresentano una classe di modelli per valutazione delle prestazioni dei sistemi di produzione che forniscono intuizioni utili sul comportamento del sistema, ma spesso con ipotesi restrittive. Quando i modelli continui vengono utilizzati per la valutazione delle prestazioni della produzione discreta sistemi, potrebbero introdurre inesattezze significative. Questo lavoro presenta un metodo analitico continuo approssimato stocastico per valutare le prestazioni dei sistemi di produzione discreti asincroni. La novità consiste nell'integrazione di meccanismi di controllo che consentono la modellazione delle caratteristiche peculiari dei sistemi di produzione discreti nel dominio continuo. Il modello si basa su una rappresentazione tramite catena Markov a tempo continuo e a stati misti continui e discreti. Un nuovo set di equazioni per valutare la propagazione dei fenomeni a livello di sistema è fornito. La validazione numerica mostra buoni risultati per il modello confrontato con la simulazione di eventi discreti. L'applicabilità in un contesto industriale è dimostrata attraverso l'analisi di due casi reali. Il primo caso include la valutazione di una politica di manutenzione opportunistica in una produzione linea che produce componenti per mobili pronti da montare. Il secondo caso analizza politiche di restart in una linea di produzione di biscotti, caratterizzata da vincoli tecnologici di processo e scarti, con l'obiettivo di migliorare la qualità della produzione.
Stochastic approximate analytical methods integrating control mechanisms for evaluating the performance of asynchronous manufacturing systems
MAGNANINI, MARIA CHIARA
Abstract
Manufacturing companies have been operating in more and more dynamic contexts, and the overall goal is represented by the ability to act proactively and reactively to unpredicted and disruptive events, at all decision levels. Model-based methods allow what-if analysis, as well as the evaluation of situations that have never been observed in practice. In the literature, stochastic approximate analytical methods represent a class of models for performance evaluation of manufacturing systems providing useful intuitions about system behavior but often with restrictive assumptions. When continuous models are used for the performance evaluation of discrete manufacturing systems, they might introduce significant inaccuracies. This work presents a stochastic approximate continuous analytical method for evaluating the performance of asynchronous discrete manufacturing systems. The novelty consists in the integration of control mechanisms that allow the modeling of the peculiar features of discrete manufacturing systems in the continuous domain. The model is based on continuous-time mixed continuous- and discrete-state Markov Chain representation. A novel set of equations to evaluate the propagation of system-level phenomena is provided. Numerical validation shows good results for the model compared to discrete-event simulation. The applicability in an industrial context is demonstrated through the analysis of two real cases. The first case includes the evaluation of an Opportunistic Maintenance policy in a manufacturing line producing components for ready-to-assemble furniture. The second case analyzes different restart policies in a food production system characterized by mixed continuous-discrete production, with the goal of improving production quality.File | Dimensione | Formato | |
---|---|---|---|
thesis.pdf
solo utenti autorizzati dal 19/07/2022
Descrizione: File tesi pdf
Dimensione
4.4 MB
Formato
Adobe PDF
|
4.4 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/10589/152297