Data-Driven Modeling and Control of Cyclic Variability of an Engine Operating in Low Temperature Combustion Modes. Issue 20 (2021)

• Record Type: Journal Article
• Authors: Batool, Sadaf; Naber, Jeffrey D.; Shahbakhti, Mahdi
• Is Part Of: IFAC-PapersOnLine. Volume 54:Issue 20(2021)
• Page Start: 834
• Page End: 839
• Publication Date: 2021
• Subjects: Combustion Stability -- Machine Learning -- Non-linear Model Predictive Control -- Homogeneous Charge Compression Ignition (HCCI) -- Reactivity Controlled Compression Ignition (RCCI); Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2021.11.275
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: Combustion cyclic variability in an internal combustion engine leads to cyclic variations in the engine torque output and emissions. Combustion cyclic variability is often characterized by coefficient of variation of indicated mean effective pressure (COVIMEP ) that is used as an indicator of combustion stability. These cyclic variations are inevitable and cannot be completely eliminated but can be controlled to allow stable engine operation. This work focuses on control oriented modeling of COVIMEP to limit engine cyclic variations in low temperature combustion (LTC) modes. COVIMEP is generally stochastic in nature; thus, a data-driven approach is used to develop a predictive model of COVIMEP for Homogeneous Charge Compression Ignition (HCCI) and Reactivity Controlled Compression Ignition (RRCI) modes. This work presents the development of a cycle-by-cycle model predictive controller for a 2.0 liter multi-mode LTC engine. Physics-based control-oriented models for combustion phasing (CA50) and IMEP are augmented with the new data-driven COVIMEP model to limit the cyclic variations below 3% for HCCI and RCCI modes. These models are then used to design closed-loop non-linear model predictive controllers to control CA50 and IMEP while constraining COVIMEP to ensure stable engine operation for varying load conditions.