Proposal for an angle/time-cyclostationary parametric model with application on internal combustion engine analysis

• Ano de defesa: 2020
• Autor(a) principal: Jambersi, Andreyson Bicudo
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Cicloestacionariedade ângulo/tempo; Análise de vibrações; Máquinas rotativas e alternativas; Motores de combustão interna; Identificação de sinais; Identificação paramétrica; Angle/time-cyclostationarity; Vibration analysis; Rotating and reciprocating machines; Internal combustion engines; Signal identification; Parametric identification
• Link de acesso: http://hdl.handle.net/11449/193465

The identification and model-based vibration analysis of a diesel internal combustion engine (ICE) is addressed in this thesis. An angle/time-dependent parametric model is proposed based on the existent well-known functional series time-varying autoregressive moving-average (FS-TARMA) formulation. However, instead of imposes the model coefficients' temporal evolution, this thesis suggests a model in which the coefficients are expanded and modeled as evolving according to an angular-periodic (related to the crankshaft position) basis. Due to this angle/time-dependency, we refer to this new model as an angle/time-FS-ARMA (or AT-FS-ARMA) model. The angular-periodic basis is derived from the AT-cyclostationary framework, a novel class of model extending the only time- or angle-cyclostationary framework. The proposed model considers the angle-periodicities (kinematic-related) often present in vibration-based signals from rotating and reciprocating machines, yet preserving the carrier (time-dependent dynamic). A ``Relaxed'' Multi-Stage Weighted-Least-Squares estimator is used to compute the models. The model structure is selected by evaluating the Bayesian information criterion through the assessment of several candidate models. A comparison between a model-based angle-frequency map and its non-parametric counterpart (obtained through smoothed-pseudo-Wigner-Ville distribution) was performed to show our proposal's potentiality. The results showed that the AT-FS-ARMA-based angle/frequency map provides a useful complementary tool for analysis; it is free from cross-terms and allows one to relate kinematics phenomena to the angular-instants in which they occur.