Magneto-rheological (MR) dampers represent one of the most promising semi-active solutions in modern automotive suspensions, owing to their extremely fast response time and continuously variable control, which allow them to modulate the dissipative action across the whole bandwidth of vehicle dynamics with limited power consumption. However, as any mechanical component, they are subject to a gradual aging process whose effects cause degradation in terms of passenger comfort and safety conditions. Therefore, aging-adaptive control methods need to be implemented to maintain such optimal conditions. While the literature extensively addresses micro-level chemistry alterations of the MR fluid after aging, the long-term degradation of the fully assembled damper, and its consequences on vehicle behavior, remains comparatively neglected. On this topics, this thesis delivers two interconnected contributions. First, it establishes an experimental procedure to track long-term degradation, by combining an aging campaign with a control-oriented Hammerstein-Wiener model, allowing to interpret parameter variations - pre and post aging- as physical signatures of aging. Second, given the experimental models of the damper, a Random Forest classifier is developed to detect the state of aging of the damper on a simulated full-car model, guaranteeing a macro-F1 score of 98.1%. Then, based on the classifier, an algorithm is designed to track the progressive aging of the dampers. The algorithm's validation is performed in simulation on realistic scenarios, such as the Worldwide Harmonized Light Vehicles Test Procedure (WLTP), guaranteeing an average estimation error of 5.4%. The outcome is data-driven diagnostic tool that effectively monitors suspension health, providing a methodical framework for future health-aware adaptive control.
Nel panorama delle moderne sospensioni automobilistiche, gli ammortizzatori magnetoreologici (MR) si collocano tra le soluzioni semi-attive di maggiore interesse grazie a tempi di risposta rapidi e a un controllo continuo, che permettono di modulare l’azione dissipativa sull’intera banda passante del veicolo limitando i consumi energetici. Tuttavia, come qualsiasi componente meccanico, i sistemi MR sono soggetti a un graduale processo di invecchiamento, con un conseguente decadimento delle prestazioni in termini di comfort e sicurezza dei passeggeri. Diventa quindi fondamentale introdurre strategie di controllo adattive per preservare le dinamiche di guida ottimali. Sebbene la letteratura analizzi ampiamente le alterazioni chimiche microscopiche del fluido MR, l'impatto della degradazione a lungo termine sull’ammortizzatore assemblato — e sul veicolo — risulta ancora poco esplorato. In questo contesto, il presente lavoro di tesi fornisce due contributi strettamente interconnessi. Il primo consiste in una metodologia sperimentale per valutare il degrado a lungo termine tramite una campagna di invecchiamento e test di caratterizzazione (pre e post-processo), finalizzati a identificare un modello di Hammerstein-Wiener orientato al controllo; in questo modo, le variazioni parametriche fungono da impronte fisiche dell'usura. Il secondo riguarda lo sviluppo di un classificatore Random Forest che, basandosi sui modelli sperimentali, rileva lo stato di invecchiamento del componente su un modello simulato di veicolo completo, raggiungendo un macro-F1 score del 98,1%. A partire da tale classificatore, è stato progettato un algoritmo per il monitoraggio continuo della degradazione, validato in simulazione su cicli standardizzati WLTP con un errore medio di stima contenuto entro il 5,4%. Il lavoro si conclude con uno strumento diagnostico data-driven in grado di monitorare lo stato di salute delle sospensioni, ponendo le basi per futuri controlli adattivi di tipo health-aware.
Degradation of Magnetorheological Dampers: experimental characterization and virtual on-vehicle health classification
Bonafini, Andrea
2025/2026
Abstract
Magneto-rheological (MR) dampers represent one of the most promising semi-active solutions in modern automotive suspensions, owing to their extremely fast response time and continuously variable control, which allow them to modulate the dissipative action across the whole bandwidth of vehicle dynamics with limited power consumption. However, as any mechanical component, they are subject to a gradual aging process whose effects cause degradation in terms of passenger comfort and safety conditions. Therefore, aging-adaptive control methods need to be implemented to maintain such optimal conditions. While the literature extensively addresses micro-level chemistry alterations of the MR fluid after aging, the long-term degradation of the fully assembled damper, and its consequences on vehicle behavior, remains comparatively neglected. On this topics, this thesis delivers two interconnected contributions. First, it establishes an experimental procedure to track long-term degradation, by combining an aging campaign with a control-oriented Hammerstein-Wiener model, allowing to interpret parameter variations - pre and post aging- as physical signatures of aging. Second, given the experimental models of the damper, a Random Forest classifier is developed to detect the state of aging of the damper on a simulated full-car model, guaranteeing a macro-F1 score of 98.1%. Then, based on the classifier, an algorithm is designed to track the progressive aging of the dampers. The algorithm's validation is performed in simulation on realistic scenarios, such as the Worldwide Harmonized Light Vehicles Test Procedure (WLTP), guaranteeing an average estimation error of 5.4%. The outcome is data-driven diagnostic tool that effectively monitors suspension health, providing a methodical framework for future health-aware adaptive control.| File | Dimensione | Formato | |
|---|---|---|---|
|
[Bonafini]_Executive_Summary.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
1.07 MB
Formato
Adobe PDF
|
1.07 MB | Adobe PDF | Visualizza/Apri |
|
[Bonafini]_Master_Thesis.pdf
non accessibile
Descrizione: Tesi
Dimensione
29.51 MB
Formato
Adobe PDF
|
29.51 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/10589/261320