The present thesis addresses the topic of safety on rural state roads, which have shown serious issues in terms of crashes. This brought to the urgent need of innovative approaches that allow to overcome the limitations of traditional methods based solely on observed crashes and are usually based on predictive methods. In fact, predictive models would allow road managers to identify potential infrastructure-related risks before real crashes occur. Initially, the aim is to adapt the worldwide recognized Highway Safety Manual (HSM) model to the Italian context and this showed the tendency of the American model to overpredict crashes. This might be due to the different strategies in crash reporting in the analyzed countries but also to the different road geometric characteristics (mainly due to curvature). Thus, specific adjustments are needed to enhance the model accuracy, particularly during the calibration process. Moreover, alternative models that better represent the Italian local conditions are developed, which proved to be more consistent with the analyzed context and highlighted the importance of considering additional variables neglected by HSM model, such as travel speed. The sensitivity analyses on the models allowed to assess their robustness with respect to changes in boundary conditions. Finally, the safety performance analysis is conducted by considering a set of different Safety Performance Indicators (SPIs), both based on predictive models or solely relying on observed crashes. Consistent discrepancies arise when ranking the safety of road sections according to different SPIs and solely relying on SPIs based on observed crashes provides limited insight into actual safety performance. The study also remarks the importance of combining the results from predictive models with those arising from on-site inspections, to identify the potential criticalities in the infrastructure to address them and eventually counteract them with specific interventions. The predictive methods also allow to perform simulations of future scenarios to assess the impact of variations in boundary conditions, such as traffic volume and/or speed, giving a valuable insight to road managers who need to intervene by adopting targeted interventions.
La presente tesi tratta il tema della sicurezza sulle strade statali extraurbane, che hanno mostrato seri problemi in termini di incidentalità. Questo ha condotto all’urgente bisogno di strategie innovative, che generalmente si avvalgono di metodi predittivi, in grado di superare le limitazioni tipiche dei tradizionali approcci alla sicurezza, basati solamente sugli incidenti osservati. Infatti, i metodi predittivi permettono ai gestori delle infrastrutture di indentificare potenziali rischi prima che conducano ad incidenti. Inizialmente, lo scopo è stato quello di adattare il modello Highway Safety Manual (HSM), riconosciuto al livello internazionale, al contesto italiano, ma questo ha evidenziato la tendenza del modello americano a sovrastimare gli incidenti. Questo potrebbe essere dovuto sia alle differenti modalità di registrazione degli incidenti nei due paesi, ma anche alle differenze geometriche nell’infrastruttura (principalmente in termini di curvatura). Pertanto, è stato necessario un adattamento del modello, specialmente durante il processo di calibrazione per migliorarne l’affidabilità. Inoltre, sono stati sviluppati dei modelli alternativi che meglio rappresentassero il contesto locale italiano, che hanno dimostrato l’importanza di considerare alcune variabili aggiuntive trascurate dal modello HSM, come la velocità di marcia. L’analisi di sensitività ha permesso di valutare la robustezza dei modelli rispetto alle variazioni nelle condizioni al contorno. Infine, le prestazioni di sicurezza della strada sono state valutate per mezzo di diversi indicatori di sintesi dell’incidentalità (SPI_S), basati sia sui modelli predittivi che sui soli incidenti osservati. Sono emerse sostanziali discrepanze nelle analisi di sicurezza condotte mediante differenti SPIs e fare solo riferimento a indicatori basati sugli incidenti osservati può portare a una visione fortemente limitata della realtà. Lo studio sottolinea anche l'importanza di combinare i risultati dei modelli predittivi con delle ispezioni in sito, che permettano di identificare possibili criticità nell’infrastruttura ed eventualmente affrontarle mediante specifici interventi. I metodi predittivi permettono inoltre di simulare scenari futuri che prevedano cambiamenti in importanti variabili come il volume di traffico o la velocità, fornendo un utile strumento decisionale ai gestori della rete stradale.
Road safety on italian state roads through predictive approach: transferability of existing models and development of new models
ANTONIAZZI, ARIANNA
2024/2025
Abstract
The present thesis addresses the topic of safety on rural state roads, which have shown serious issues in terms of crashes. This brought to the urgent need of innovative approaches that allow to overcome the limitations of traditional methods based solely on observed crashes and are usually based on predictive methods. In fact, predictive models would allow road managers to identify potential infrastructure-related risks before real crashes occur. Initially, the aim is to adapt the worldwide recognized Highway Safety Manual (HSM) model to the Italian context and this showed the tendency of the American model to overpredict crashes. This might be due to the different strategies in crash reporting in the analyzed countries but also to the different road geometric characteristics (mainly due to curvature). Thus, specific adjustments are needed to enhance the model accuracy, particularly during the calibration process. Moreover, alternative models that better represent the Italian local conditions are developed, which proved to be more consistent with the analyzed context and highlighted the importance of considering additional variables neglected by HSM model, such as travel speed. The sensitivity analyses on the models allowed to assess their robustness with respect to changes in boundary conditions. Finally, the safety performance analysis is conducted by considering a set of different Safety Performance Indicators (SPIs), both based on predictive models or solely relying on observed crashes. Consistent discrepancies arise when ranking the safety of road sections according to different SPIs and solely relying on SPIs based on observed crashes provides limited insight into actual safety performance. The study also remarks the importance of combining the results from predictive models with those arising from on-site inspections, to identify the potential criticalities in the infrastructure to address them and eventually counteract them with specific interventions. The predictive methods also allow to perform simulations of future scenarios to assess the impact of variations in boundary conditions, such as traffic volume and/or speed, giving a valuable insight to road managers who need to intervene by adopting targeted interventions.File | Dimensione | Formato | |
---|---|---|---|
2025_04_Antoniazzi.pdf
solo utenti autorizzati a partire dal 22/04/2026
Dimensione
38.03 MB
Formato
Adobe PDF
|
38.03 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/10589/237897