Most developed countries worldwide are dealing with large stocks of existing bridges approaching the end of the service life. The preservation of the structural and functional adequacy of these bridges is a priority for administrations and decision-makers. Visual inspections are at the base of an effective and reliable bridge condition assessment. The collection of information over time provides a great amount of bridge data which can be properly elaborated to get useful insights for supporting the decision-making process. In this context, the analysis of massive amount of data included in digital bridge inventories through proper data Machine Learning (ML) tools can be exploited to support the decision-making process and to prioritize maintenance interventions within the transportation network. The thesis proposes the application of different ML algorithms for the elaboration of bridge data included in the California National Bridge Inventory (NBI) aimed at the condition assessment of bridge components. The effectiveness and interpretability of the proposed models are finally discussed.
Il patrimonio infrastrutturale di molti Paesi è caratterizzato da numerosi manufatti che hanno ormai raggiunto la vita utile per la quale sono stati progettati. Pertanto, molte pubbliche amministrazioni e enti gestori hanno l’urgenza di definire le priorità di intervento. Le ispezioni visive consentono di ricavare una grande quantità di informazioni essenziali per una corretta valutazione dello stato di conservazione e un’ottimale pianificazione degli interventi di manutenzione. In questo contesto, gli strumenti di Machine Learning (ML) sono particolarmente adatti all’analisi delle grandi quantità di dati raccolti in database informatizzati e possono essere di supporto nel processo decisionale. La tesi propone l’applicazione di alcuni algoritmi di ML per l'elaborazione delle informazioni relative ai ponti in California e archiviati nel National Bridge Inventory (NBI) al fine di valutare la condizione dei componenti del ponte. Viene infine discussa l'efficacia e l'interpretabilità dei modelli proposti.
Condition rating for bridge components using machine learning methods
Lu, Tian
2021/2022
Abstract
Most developed countries worldwide are dealing with large stocks of existing bridges approaching the end of the service life. The preservation of the structural and functional adequacy of these bridges is a priority for administrations and decision-makers. Visual inspections are at the base of an effective and reliable bridge condition assessment. The collection of information over time provides a great amount of bridge data which can be properly elaborated to get useful insights for supporting the decision-making process. In this context, the analysis of massive amount of data included in digital bridge inventories through proper data Machine Learning (ML) tools can be exploited to support the decision-making process and to prioritize maintenance interventions within the transportation network. The thesis proposes the application of different ML algorithms for the elaboration of bridge data included in the California National Bridge Inventory (NBI) aimed at the condition assessment of bridge components. The effectiveness and interpretability of the proposed models are finally discussed.File | Dimensione | Formato | |
---|---|---|---|
Master_Thesis_Tian_LU.pdf
non accessibile
Dimensione
34.45 MB
Formato
Adobe PDF
|
34.45 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/10589/189738