Robust Estimation of Fuel Pump Volumetric Efficiency Using Tree-Based Ensembles
Küçük Resim Yok
Tarih
2025
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Institute of Electrical and Electronics Engineers Inc.
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
Volumetric efficiency (VE) measurement in highpressure fuel pumps, specifically at 35 MPa (LG35), is critical for diagnosing leakage and wear. To analyze pressure traces, conventional methods rely on manual analysis, which is timeconsuming and requires expert intervention. We propose a machine learning-based alternative that estimates LG35 from routinely logged cycle-level metrics such as stroke length, shaft speed, system pressure, and temperature. Using a dataset of 4,139 real-world pump cycles, we benchmark seven regression models and demonstrate that tree-based ensembles, particularly Gradient Boosting, significantly outperform linear baselines. A voting ensemble combining multiple models achieves a mean absolute error of 3.04 % on a hold-out set simulating unseen stroke conditions. SHAP analysis confirms that model predictions align with physical intuition. Our approach enables real-time, expert-free diagnostics, offering a scalable solution for fuel system testing in automotive manufacturing. © 2025 IEEE.
Açıklama
2025 Innovations in Intelligent Systems and Applications Conference, ASYU 2025 -- 2025-09-10 through 2025-09-12 -- Bursa -- 214381
Anahtar Kelimeler
ensemble learning, fuel pump testing, gradient boosting, regression models, Volumetric efficiency
Kaynak
WoS Q Değeri
Scopus Q Değeri
N/A
