Control-Oriented Modeling and Model-Based Control of a Blowdown Supercharge (BDSC) HCCI Engine

Jagsch Michael, Yasuo Moriyoshi, Tatsuya Kuboyama
Closed-loop control is vital for an application of HCCI engines in passenger cars. This paper introduces a simplified control-oriented model for control of combustion phasing and IMEP of a Blowdown Supercharge Engine (BDSC). Despite the complexity of this particular engine, the model has been found to match not only the steady state values in high load HCCI, but also to reproduce the transients. This model takes advantage of the knowledge of non-dynamic processes within the engine that can be
more » ... ngine that can be derived from steady state values, while the main dynamics are achieved by dynamically modelling of cyclic coupling via in-cylinder temperature alone and mean exhaust pressure dynamics. Furthermore, a simplified combustion model has been found to be accurate enough for the region of interest. An automated tuning scheme helps to match the model to the respective target values. With this model and the tuning scheme, the model can be easily tuned for every possible case. A model-based MIMO state controller, based on Sliding-Mode Control theory has been designed and tested on a detailed 1-D simulation code. KEY WORDS: heat engine, homogeneous charge compression ignition, theory/modeling, design/control [A1] 1.Introduction Stricter emission regulations and an increasing demand in fuel economy, together with the consideration of alternative propulsion methods, such as the recent electrification of cars, impose a great pressure on the improvement of nowadays combustion engines. While traditionally gasoline engines feature relatively low emissions, and diesel concepts offer in relation low fuel consumption, the Homogeneous Charge Compression Ignition (HCCI) concept has the potential to combine both advantages within one single concept. Due to that, extensive research has been undertaken, and their application in future passenger cars is considered possible. Despite these advantages, unresolved issues still hinder these engines from the implementation in cars. One of the main drawbacks is the conception that HCCI is difficult to control, since a direct actuator for combustion does not exist. Several control strategies for ignition actuation has been researched extensively in the past. Additionally there exists the common agreement that HCCI cannot cover all necessary operating points and hence cannot replace a conventional combustion engine. Therefore, HCCI has to be operated in tandem with SI or diesel, where switching between conventional combustion and HCCI becomes necessary. Due to the sensitivity of combustion on in-cylinder conditions, which in turn depend on the permanently varying atmospheric conditions within an entire lifecycle of a car, combustion needs to be made robust by closed-loop control. The Blowdown Supercharge (BDSC) variant of HCCI has been intensively studied at Chiba University and several measures have been carried out to increase the operational range of HCCI. One of the main concepts is the use of a blowdown pressure wave in the exhaust pipe between two connected cylinders that boosts additional exhaust gases into one of the cylinder, from the pressure wave of the other cylinder, and this concept has been first introduced by Hatamura (1) . Several closed-loop control schemes for other and conceptually simpler versions of HCCI have been presented in the past (Johansson (2) et al., Ravi (3) et al., Widd (4) et al., Killingsworth (5) et al., Rausen (6) et al.), but the higher complexity of the BDSC engine with its boosting pressure wave in the exhaust port, and thereby the expected higher modeling effort, has prevented a model-based control scheme so far. the temperature and pressure at start, and T 2 and P 2 the values at end compression or expansion, V 1 and V 2 their respective and known volumes, and γ as the polytropic exponent. For prediction of start of combustion, the ignition model introduced above can be employed in the control-oriented model as well. Since however, the polytropic exponent remains constant throughout compression, the ignition model from the detailed BOOST model would not re-produce the same results and the accuracy of the ignition model is not guaranteed any more. Even more, since there is no dynamic interaction between the gases and the cylinder walls during compression in the model, the entire sequence of temperature and pressure is already known at start of compression. Therefore, the ignition model has to be amended and due to the predictability of ignition timing at start of compression, a different approach has been pursued here. The underlying idea is to convert the ignition model, as presented above, to a modified ignition delay model that estimates the start of combustion, by only knowing the initial conditions at start of compression. The selected form resembles an Arrhenius function, as well as the ignition delay above. While the ignition delay function above has been measured and fitted under constant temperature conditions, the temperature is changing here due to the ongoing compression of the in-cylinder gases by the piston. Hence, the activation energy has to be expected lower than in the original ignition model. In addition to that, in Chiang (14,15) et al., as well as in Jagsch (7) , the dominance of the temperature has been pointed out, and it was concluded that good results can be obtained by neglecting the gas composition. A correlation that is simple but precise enough has been found of the following form: A is hereby a pre-exponential factor that in this case may compensate for the omitted gas composition terms, E as a modified activation energy in order to compensate for the gradual temperature change of the upward moving piston and C an offset to account for combustion phasing around 360 deg CAD. The exponent B has been set to a fixed negative value. In fact, it was found that optimizing B would lead to values close to zero, which implies that the pressure term could be omitted. A least-squares algorithm has been used here, and the results are displayed in Figure 7 , with CA50 as the target, of which the calculated values show a superb fit to the target values in a particular wide range, with non-normalized residuals less than one degree CAD. The reference points were picked randomly from the BOOST model, providing a wide range of cases to tune the overall model, with IMEP between 241.23 kPa and 536.22 kPa" CA50 BD 
doi:10.20485/jsaeijae.5.4_145 fatcat:5h4ekb6ignho5nis6buykxsh4y