Anders Widd

Physics-Based Model Predictive Control of HCCI Combustion Phasing Using Fast Thermal Management and VVA

Homogeneous charge compression ignition (HCCI) is a promising internal combustion engine concept. It holds promise of combining low emission levels with high efficiency. However, as ignition timing in HCCI operation lacks direct actuation and is highly sensitive to operating conditions and disturbances, robust closed-loop control is necessary. To facilitate control design and allow for porting of both models and the resulting controllers between different engines, physics-based mathematical models of HCCI are of interest. This paper presents work on a physical model of HCCI including cylinder wall temperature and evaluates predictive controllers based on linearizations of the model. The model was derived using first principles and formulated on a cycle-to-cycle basis. The resulting model was of second order with two inputs and two outputs. Measurement data including cylinder wall temperature measurements was used for calibration and validation of the model. Predictive control of the combustion phasing was then evaluated experimentally using ethanol as fuel. The control signals were the intake temperature and the inlet valve closing timing. The control performance was evaluated in terms of response time and steady-state output variance. Multi-cylinder control experiments were also carried out.

A Physical Two-Zone NOx Model Intended for Embedded Implementation

This paper offers a two-zone NOx model suitable for vehicle on-board, on-line implementation. Similar NOx modeling attempts have previously been undertaken. The hereby suggested method does however offer clear and important benefits over the previously methods, utilizing a significantly different method to handle temperature calculations within the (two) different zones avoiding iterative computation. The new method significantly improves calculation speed and, most important of all, reduces implementation complexity while still maintaining reasonable accuracy and the physical interpretation of earlier suggested methods. The equations commonly used to compute NOx emissions is also rewritten in order to suit a two-zone NOx model. An algorithm which can be used to compute NOx emissions is presented and the intended contribution of the paper is a NOx model, implementation feasible for an embedded system, e.g. embedded processor or embedded electronic hardware (FPGA). For that purpose parts of the algorithm can be pre-computed and stored in tables allowing significant acceleration of the computation.

Experimental evaluation of predictive combustion phasing control in an HCCI engine using fast thermal management and VVA

This paper presents experimental results on model predictive control of the combustion phasing in a Homogeneous Charge Compression Ignition (HCCI) engine. The controllers were based on linearizations of a previously presented physical model of HCCI including cylinder wall temperature dynamics. The control signals were the inlet air temperature and the inlet valve closing. A system for fast thermal management was installed and controlled using mid-ranging control. The resulting control performance was experimentally evaluated in terms of response time and steady-state output variance. For a given operating point, a comparable decrease in steadystate output variance was obtained either by introducing a disturbance model or by changing linearization point. The robustness towards disturbances was investigated as well as the effects of varying the prediction and control horizons.

Control of exhaust recompression HCCI using hybrid model predictive control

Homogeneous Charge Compression Ignition (HCCI) holds promise for reduced emissions and increased efficiency compared to conventional internal combustion engines. As HCCI lacks direct actuation over the combustion phasing, much work has been devoted to designing controllers capable of set-point tracking and disturbance rejection. This paper presents results on model predictive control (MPC) of the combustion phasing in an HCCI engine based on a hybrid model formulation composed of several linearizations of a physics-based nonlinear model. The explicit representation of the MPC was implemented experimentally and the performance during set point changes was compared to that of a switched state feedback controller. The hybrid MPC produced smoother transients without overshoot when the set point change traversed several linearizations.

Controlling Combustion Phasing of Recompression HCCI with a Switching Controller

Homogeneous charge compression ignition (HCCI) is more efficient and produces significantly less NOx emissions compared to spark ignitions. Using an exhaust recompression strategy to achieve HCCI, however, produces cycle-to-cycle coupling which makes the problem of controlling combustion phasing more difficult. In the past, a linear feedback controller designed with a single linearized model is effective in controlling combustion phasing around an operating point. However, HCCI dynamics can change dramatically around different operating points such that a single linearization is insufficient to approximate the entire operating range. Further investigation shows that the operating range can be roughly divided into three regions where a linear model can capture the qualitative system behavior in each of the regions. As a result, a three zone switching linear model approximates recompression HCCI dynamics far better than a single linearization. This new model structure also suggests that two of the three regions need completely opposite control actions. Therefore, the approach of using a static feedback control based on a single linearziation cannot be appropriate over the entire operating range. We propose a switching controller based on the switching linear model and achieve very good performance in controlling HCCI combustion phasing throughout the entire operating region. Lastly, a semi-definite programming (SDP) formulation of finding a Lyapunov function for the switching linear model is presented in order to guarantee stability of the switching control scheme.

Steady state fuel consumption optimization through feedback control of estimated cylinder individual efficiency

Engine efficiency is often controlled in an indirect way through combustion timing control. This requires a priori knowledge of where to phase the combustion for different operating points and conditions. With cylinder individual efficiency estimation, control strategies aiming directly at fuel consumption optimization can be developed. It has previously been shown that indicated efficiency can be estimated using the cylinder pressure trace. This paper presents a method to use the estimated efficiency as a feedback variable in an extremum seeking control strategy for online steady state fuel consumption optimization. The experimental results show that the controller manages to find the maximum brake torque region at the given operating point both with and without external excitation.

Investigating Mode Switch from SI to HCCI using Early Intake Valve Closing and Negative Valve Overlap

This study investigates mode switching from spark ignited operation with early intake valve closing to residual gas enhanced HCCI using negative valve overlap on a port-fuel injected light-duty diesel engine. A mode switch is demonstrated at 3.5 bar IMEPnet and 1500 rpm. Valve timings and fuel amount have to be selected carefully prior to the mode switch. During mode transition, IMEPnet deviates by up to 0.5 bar from the set point. The time required to return to the set point as well as the transient behavior of the engine load varies depending on which control structure that is used. Both a model-based controller and a PI control approach were implemented and evaluated in experiments. The controllers were active in HCCI mode. The model-based controller achieved a smoother transition and while using it, the transition could be accomplished within three engine cycles. The initial deviation in load is unacceptable but can most likely be improved with a predictive mode transition model compared to empirically selected mode transition parameters. Changing the fuel injection method to direct injection instead of port injection is another possible improvement. (Less)

Single-zone Diesel PPC modeling for control

Partially premixed combustion (PPC) is a combustion concept with similarities to both Diesel and Homogeneous Charge Compression Ignition (HCCI) combustion. It provides a combustion mode with better controllability than HCCI without increasing the emissions of nitrogen oxides and soot to the level of traditional Diesel engines. The model described in this paper aims to describe the main features of Diesel PPC combustion within the closed part of an engine cycle. It is a single-zone model including heat losses to the cylinder walls as well as fuel evaporation losses. The simulation framework used allows optimization problems to be formulated based on the model equations. The overarching goal of this modeling is to be able to use the model explicitly for optimization.

A Fast Physical NOx Model Implemented on an Embedded System

This paper offers a two-zone, physical, NOx model with low computational cost, implemented in C on an embedded system. The model is able to compute NOx-emission formation with high time resolution during an engine cycle. To do this the model takes cylinder pressure and injected fuel amount as inputs and produces NO concentration as output. The model as such is not new, nevertheless the physical background of the model as well as the equations upon which the model is based had to be briefly described to facilitate the understanding of the subsequent work. The main part of the paper is devoted to the process of developing an algorithm implementing the described model, techniques used and issues encountered are described. The resulting algorithm was implemented in C and tested on an embedded ARM processor. For the sake of implementation, parts of the algorithm had to be pre-computed and stored in tables, allowing significant acceleration of the computations. Since the model is non-linear, exponentially spaced tables had to be developed in order to successfully tabulate the parts needed without consuming too much memory. Much of the methods presented are also applicable in a variety other applications when it is desirable to implement fast versions of complex algorithms and models. The outcome regarding computation speed and memory needed is discussed. The final result is a low-cost NOx model, which is able to compute several orders of magnitude faster than NOx models known so far, implemented in C on an embedded system.

Modeling for HCCI Control

Due to the possibility of increased efficiency and reduced emissions, Homogeneous Charge Compression Ignition (HCCI) is a promising alternative to conventional internal combustion engines. Ignition timing in HCCI is highly sensitive to operating conditions and lacks direct actuation, making it a challenging subject for closed-loop control. This paper presents physics-based, control-oriented modeling of HCCI including cylinder wall temperature dynamics. The model was calibrated using experimental data from an optical engine allowing measurements of the cylinder wall temperature to be made. To further validate the model, it was calibrated against a conventional engine and linearizations of the model were used to design model predictive controllers for control of the ignition timing using the inlet valve closing and the intake temperature as control signals. The resulting control performance was experimentally evaluated in terms of response time and steady-state output variance.