The management of military fleets represents a complex problem, that encompasses the creation of an annual flight plan for aircraft in alignment with national and international missions, as well as scheduling maintenance considering the limited capacity of maintenance docks and varying types of interventions. This is done while accounting for diverse requirements based on aircraft types and the fleet to which they belong. Starting from mixed-integer optimization problem previously formulated to address fleet management under sub-optimal initial conditions, a strategy was introduced to ensure that an optimal management plan is found even in the extreme cases of fleets with unbalanced initial conditions, which was instead not possible with the original formulation. The introduction of this strategy required the development of an automated process capable of finding the optimal set of hyperparameters of the optimization problem to achieve the predefined objectives. Bayesian optimization, utilized for the hyperparameter selection process, significantly reduced the time required for model calibration, making the tool user-friendly even for military personnel less familiar with the specific optimization problem considered in this thesis. The innovative strategies introduced were subsequently tested under various initial conditions, and their results were compared with those previously available. The strategies presented in this thesis offer increased adaptability of the model to the fleet's initial conditions, aiding the military aviation in the digitalization of the fleet management process.
La gestione di flotte militari rappresenta un problema complesso, che comprende la stesusa del piano di volo annuale dei velivo in accordo con le missioni nazionali e internazioni, la schedulazione delle manutenzioni considerando la ridotta capacità delle stazioni manutentive e i differenti tipi di interventi. Tenendo in considerazion differenti esigenze in funzione del tipo di velivoli e della flotta a cui appartine. A partire dal problema di ottimizzazione misto-intero formulato in precedenza per affrontare la gestione delle flotte in condizioni iniziali subottimali si è introdota una strategia in grado di garantire la risoluzione anche nei casi limite di flotte con condizioni iniziali completamente sbilanciate. L'introduzione della strategia ha reso necessario lo sviluppo di un processo automatico in grado di trovare il miglior set di iperparametri del modello per il raggiungimento degli obbiettivi preposti. L'ottimizzazione Bayesiana sfruttata per il processo di scelta degli iperparametri ha permesso una rilevante riduzione del tempo necessario alla calibrazione del modello rendendo inoltre lo strumento di facile utilizzo anche al personale militare meno affine con il problema di ottimizzazione considerato in questa tesi. Le strategie innovative introdotte sono quindi state testate con differenti condizioni iniziali andando a comparare i risultati con quelli precedentemente disponibili. Le startegie introdotte in questa tesi forniscono maggiore adattabilità del modello alle condizioni iniziali della flotta aiutando l'aeronautica militare nella digitalizzazione del processo di gestione della flotta.
Automated scheduling of the Italian Air Force fleet via dynamic optimization with Bayesian tuning
GHISALBERTI, LORENZO
2022/2023
Abstract
The management of military fleets represents a complex problem, that encompasses the creation of an annual flight plan for aircraft in alignment with national and international missions, as well as scheduling maintenance considering the limited capacity of maintenance docks and varying types of interventions. This is done while accounting for diverse requirements based on aircraft types and the fleet to which they belong. Starting from mixed-integer optimization problem previously formulated to address fleet management under sub-optimal initial conditions, a strategy was introduced to ensure that an optimal management plan is found even in the extreme cases of fleets with unbalanced initial conditions, which was instead not possible with the original formulation. The introduction of this strategy required the development of an automated process capable of finding the optimal set of hyperparameters of the optimization problem to achieve the predefined objectives. Bayesian optimization, utilized for the hyperparameter selection process, significantly reduced the time required for model calibration, making the tool user-friendly even for military personnel less familiar with the specific optimization problem considered in this thesis. The innovative strategies introduced were subsequently tested under various initial conditions, and their results were compared with those previously available. The strategies presented in this thesis offer increased adaptability of the model to the fleet's initial conditions, aiding the military aviation in the digitalization of the fleet management process.File | Dimensione | Formato | |
---|---|---|---|
2023_10_ghisalberti.pdf
non accessibile
Descrizione: Thesis
Dimensione
10.79 MB
Formato
Adobe PDF
|
10.79 MB | Adobe PDF | Visualizza/Apri |
EA_2023_10_ghisalberti.pdf
non accessibile
Descrizione: Extended abstract
Dimensione
1.33 MB
Formato
Adobe PDF
|
1.33 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/10589/209767