The issue of heavy vehicles rollover appears to be central in various sectors. This is due to the consequences entailed in terms of driver and passenger safety, other than considering aspects as environmental damaging and pollution. Therefore, several studies proposed estimative and predictive techniques to avoid this critical condition, with especially good results obtained by the using of Artificial Intelligence (AI) systems based on neural networks. Unfortunately, to conduct these kind of analyses a great quantity of data is required, with the same often difficult to be retrieved in sufficient numbers without incurring in unsustainable costs. To answer the problem, in this paper is presented a methodology based on synthetic data, generated in a specifically designed Matlab environment. This has been done by defining the characteristics of an heavy vehicle, a three axles truck, and making it complete maneuvers on surfaces and circuits purposely created to highlight rollover issues. After this phase, represented by the generation and processing of the data, follows the analysis of the same. This is the second major phase of the methodology, and contains the definition of a neural networks based algorithm. Referring to the nets, these are designed to obtain both the estimate and the prediction of four common rollover indexes, the roll angle and the Load Transfer Ratios (LTR, one for each axle). Very promising results were achieved in particular for the estimative part, offering new possibilities for the analysis of rollover issues both for the generation and the analysis of the data.

Analysis of Heavy Vehicles Rollover with Artificial Intelligence Techniques / Velardocchia, Filippo; Perboli, Guido; Vigliani, Alessandro. - STAMPA. - 13810:(2023), pp. 294-308. (Intervento presentato al convegno 8th International Conference, LOD 2022 tenutosi a Certosa di Pontignano nel September 18–22, 2022,) [10.1007/978-3-031-25599-1_22].

Analysis of Heavy Vehicles Rollover with Artificial Intelligence Techniques

Filippo Velardocchia;Guido Perboli;Alessandro Vigliani
2023

Abstract

The issue of heavy vehicles rollover appears to be central in various sectors. This is due to the consequences entailed in terms of driver and passenger safety, other than considering aspects as environmental damaging and pollution. Therefore, several studies proposed estimative and predictive techniques to avoid this critical condition, with especially good results obtained by the using of Artificial Intelligence (AI) systems based on neural networks. Unfortunately, to conduct these kind of analyses a great quantity of data is required, with the same often difficult to be retrieved in sufficient numbers without incurring in unsustainable costs. To answer the problem, in this paper is presented a methodology based on synthetic data, generated in a specifically designed Matlab environment. This has been done by defining the characteristics of an heavy vehicle, a three axles truck, and making it complete maneuvers on surfaces and circuits purposely created to highlight rollover issues. After this phase, represented by the generation and processing of the data, follows the analysis of the same. This is the second major phase of the methodology, and contains the definition of a neural networks based algorithm. Referring to the nets, these are designed to obtain both the estimate and the prediction of four common rollover indexes, the roll angle and the Load Transfer Ratios (LTR, one for each axle). Very promising results were achieved in particular for the estimative part, offering new possibilities for the analysis of rollover issues both for the generation and the analysis of the data.
2023
978-3-031-25598-4
978-3-031-25599-1
File in questo prodotto:
File Dimensione Formato  
main.pdf

Open Access dal 10/03/2024

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 796.09 kB
Formato Adobe PDF
796.09 kB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2976669