Physically-based modeling, estimation, and control of piezoelectric fuel injection during rate shaping operation

Dat D Le, Purdue University

Abstract

Diesel engines are used commonly for heavy-duty vehicles as a result of their high power density and efficiency characteristics. However, one issue with diesel engines is NOx, particulate matter (PM), and noise emissions. Besides the effort for exhaust aftertreatment, one strategy to lower the emissions, as well as reduce overall fuel consumption, is to improve the combustion process by utilizing complex injection rate profiles, e.g. rate shaping. This is due to the fact that engine efficiency and harmful emissions depend on the mixing of the mixture between air and fuel. To illustrate, rate shaping can be used to reduce the combustion rate, enabling low temperature combustion, reducing the peak combustion temperature, and resulting in reduced NOx formation. Traditional common rail fuel systems are used with solenoid injectors. Recently, piezoelectrically-actuated fuel injectors, which have a faster response, have been developed. Piezoelectric fuel injectors deliver fuel more accurately, enabling closely spaced, multiple-pulse injection events each cycle as well as rate shaping. To date, injection rate shaping has been implemented mostly by pressure modulation or by open-loop modulation of the needle lift. In this work, a nonlinear simulation model for single, multiple pulse, and rate-shaped profiles was developed, and experimentally-validated at different operating conditions. Consideration of the flow rate nonlinearity, piezostack hysteresis, and hydro-mechanical dynamics was required to achieve precise modeling of the piezoelectric fuel injector during rate shaping operation. The simulation model was capable of capturing the rate shapes, and showed the modeling errors in injected fuel amount to less than 5.2% at 1000 bar, 1200 bar, and 1400 bar rail pressures. The simulation model was then simplified to obtain a reduced-order model. This reduced-order model was used to develop a real-time estimator, which estimates the states of the injector model during rate shaping. The estimator includes different sampling rates for different loops run in parallel on an NI CompactRIO FPGA. The estimator, which is based on the dynamics of the injector, runs at a sampling period of 8 μs. The hysteresis model loop runs at a sampling period of 2 μs. The estimator showed the predicted errors in injected fuel amount to no more than 3% at 500 bar and 600 bar rail pressures. A model-based closed-loop controller for rate shaping was developed. The control is nonlinear and based on dynamic surface control for non-smooth systems to compensate for the injector nonlinearities. The closed-loop control system was shown to be stable and achieve arbitrarily small tracking errors. The performance of this controller was verified with simulation and experiment. The controller was capable of tracking desired boot-shaped injection profiles. The controller showed the tracking errors in injected fuel amount to no more than 2.5% at 500 bar and 600 bar rail pressures.

Degree

Ph.D.

Advisors

Shaver, Purdue University.

Subject Area

Mechanical engineering

COinS