This thesis analyzes a detection error that occurred during the End-of-Line Noise & Vibration (NVH) test of some manual transmissions produced at the Stellantis plant in Verrone. Some noisy gearboxes were classified as compliant by the automated quality control test bench, revealing limitations in the order-track signal threshold used to assess the conformity of the 1st gear of the gearbox. Through the Six Sigma DMAIC methodology, the study identifies the root cause of the error, showing that the adaptive threshold, developed by Discom, i.e., the manufacturer of the NVH quality test bench, which is based on mean and standard deviation statistical metrics, is sensitive to outliers and to changes in the dimension of the statistical sample. Two alternative thresholding strategies were developed: an updated version of the threshold currently used and a new formulation based on the Median Absolute Deviation (MAD). The parameters of the latter were selected by analysing the trade-off between false rejects (good gearboxes incorrectly filtered out) and false accepts (noisy gearboxes that are not detected). The new threshold was then validated using historical production data, showing better stability over time and a reduction in missed detections compared to the previous quality assessment logic. The results demonstrate that the MAD-based threshold significantly improves the diagnostic capability of the NVH test bench and provides a solid foundation for future extensions to other gears and test conditions.
Questa tesi analizza un errore di rilevazione verificatosi nel test End-of-Line Noise & Vibration (NVH) di alcuni cambi manuali prodotti nello stabilimento Stellantis di Verrone. Alcuni cambi rumorosi erano stati classificati come conformi dal banco prova, evidenziando dei limiti della soglia del segnale di order track utilizzata per stabilire la conformità della 1ª marcia. Attraverso la metodologia Six Sigma DMAIC, lo studio individua la causa principale dell’errore, mostrando come la soglia adattiva, proposta da Discom, l’azienda produttrice del banco di prova, basata su media e deviazione standard risulti sensibile agli outlier e ai cambiamenti della dimensionalità del campione statistico sulla basa del quale questi parametri vengono calcolati. Sono state sviluppate due strategie alternative per la definizione della soglia: una versione aggiornata della soglia attualmente in uso, e una nuova formulazione basata sulla Median Absolute Deviation (MAD). I parametri di quest’ultima sono stati scelti analizzando il compromesso tra falsi scarti (cambi buoni che vengono erroneamente filtrati) e falsi OK (cambi rumorosi che non vengono individuati). La nuova soglia è stata successivamente validata utilizzando dati storici di produzione, mostrando una maggiore stabilità nel tempo e una riduzione del numero di difetti non rilevati rispetto alla precedente logica di controllo. I risultati dimostrano che la soglia basata sulla MAD migliora in modo significativo la capacità diagnostica del banco NVH e costituisce una base solida per possibili estensioni future ad altre marce e ad altre condizioni di test.
NVH threshold optimization for end-of-line Gearbox testing at Stellantis powertrain plant
Allegretti, Francesco
2024/2025
Abstract
This thesis analyzes a detection error that occurred during the End-of-Line Noise & Vibration (NVH) test of some manual transmissions produced at the Stellantis plant in Verrone. Some noisy gearboxes were classified as compliant by the automated quality control test bench, revealing limitations in the order-track signal threshold used to assess the conformity of the 1st gear of the gearbox. Through the Six Sigma DMAIC methodology, the study identifies the root cause of the error, showing that the adaptive threshold, developed by Discom, i.e., the manufacturer of the NVH quality test bench, which is based on mean and standard deviation statistical metrics, is sensitive to outliers and to changes in the dimension of the statistical sample. Two alternative thresholding strategies were developed: an updated version of the threshold currently used and a new formulation based on the Median Absolute Deviation (MAD). The parameters of the latter were selected by analysing the trade-off between false rejects (good gearboxes incorrectly filtered out) and false accepts (noisy gearboxes that are not detected). The new threshold was then validated using historical production data, showing better stability over time and a reduction in missed detections compared to the previous quality assessment logic. The results demonstrate that the MAD-based threshold significantly improves the diagnostic capability of the NVH test bench and provides a solid foundation for future extensions to other gears and test conditions.| File | Dimensione | Formato | |
|---|---|---|---|
|
2025_12_Allegretti.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
4.48 MB
Formato
Adobe PDF
|
4.48 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/10589/247433