Thivaharan Albin

Nonlinear MPC for a two-stage turbocharged gasoline engine airpath

Innovative charging concepts such as two-stage turbocharging for gasoline engines, cause high demands on the process control due to the complex, nonlinear system behavior. For complex, nonlinear systems Nonlinear Model-based Predictive Controllers (NMPC) offer a high potential. They are capable of handling coupled multiple-input systems while achieving high control quality and respecting constraints of the system. In the case of turbocharging, considerations to protect components can introduce the necessity to constrain certain system values. This paper presents a two-stage turbocharged gasoline airpath modeling approach which is suited to be used in a NMPC implementation. The control implementation is based on direct optimal control using an online Sequential Quadratic Programming (SQP) type algorithm. For validating the control performance, simulations are conducted. The computation time of the algorithm is determined by implementation on a control prototyping platform for validation of the real-time capability.

A hybrid control approach for low temperature combustion engine control

In this paper, a hybrid control approach for low temperature combustion engines is presented. The identification as well as the controller design are demonstrated. In order to identify piecewise affine models, we propose to use correlation clustering algorithms, which are developed and used in the field of data mining. We outline the identification of the low temperature combustion engine from measurement data based on correlation clustering. The output of the identified model reproduces the measurement data of the engine very well. Based on this piecewise affine model of the process, a hybrid model predictive controller is considered. It can be shown that the hybrid controller is able to produce better control results than a model predictive controller using a single linear model. The main advantage is that the hybrid controller is able to manage the system characteristics of different operating points for each prediction step.

Decoupling of consecutive gasoline controlled auto-ignition combustion cycles by field programmable gate array based real-time cylinder pressure analysis

Gasoline controlled auto-ignition combustion offers high potential for CO2 emission reduction, but faces challenges regarding combustion stability and high sensitivity to changing boundary conditions. Combustion chamber recirculation allows a wide operation range, but results in a strong coupling of consecutive cycles due to residuals that are transferred to the subsequent combustion cycle. The cycle coupling leads to phases of unstable operation with reduced efficiency and increased emission levels. State-of-the-art control algorithms use data-driven models of gasoline controlled auto-ignition combustion to achieve cycle-to-cycle control of the process or use offline calibration and optimization. A closed-loop control is proposed and implemented on a rapid control prototyping engine control unit. The control algorithm continuously calculates the current residual fuel in the combustion chamber. The heat release is observed and compared with the theoretical heat release of the injected fuel mass. The rate of unburned fuel mass transferred to the subsequent cycle is calculated offline by a detailed gas exchange model. Based on this information, the control algorithm adapts the injected fuel quantity for each cycle individually using an inverse injector model. In this article, a concept for decoupling consecutive cycles is presented to reduce the deviations of the indicated mean effective pressure and thus the heat release. Unstable sequences are analyzed in the time domain, and unburned residuals are identified as a strong correlating factor for consecutive cycles. Using real-time cylinder pressure analysis based on a field programmable gate array enables the online calculation of unburned residual fuel. Based on this calculation, the injection of each cycle can be adapted individually to decouple consecutive cycles and avoid unstable operation. The results of the control algorithm and the stabilization of the gasoline controlled auto-ignition combustion are validated using a single-cylinder research engine and compared to steady-state operation.

Model-Based Optimal Control for PCCI Combustion Engines

Abstract New combustion methods for engines have been recently researched very intensively. In diesel engines, the homogenisation of the air-fuel mixture by early fuel injection has significant effects on emission reduction. The paper presents a model-based optimal control strategy for premixed charge compression ignition (PCCI) low temperature combustion in diesel engines. In order to understand the basic properties of the PCCI mode, static and dynamic measurements were conducted using a real conventional diesel engine. The main inputs of the combustion process are the exhaust gas recirculation rate and injection parameters. Outputs are the indicated mean effective pressure and the fuel mass conversion balance point. The process has very fast, almost proportional dynamics over the engines working cycles. Focusing on the static behaviour of the process, a nonlinear neural network model is used for identification. Successive linearisation of the nonlinear network is used as predictive controller model. The presented controller structure is able to consider constraints and can be computed very fast. Finally, the controller is validated under real time conditions by experimental tests at the engine test bench. Although the controller structure contains a model and a convex optimisation step with regards to constraints, its implementation is very simple, as no observer is used, and the linearised model consists of static gains only.

Controlling GCAI (Gasoline Controlled Auto Ignition) in an Extended Operating Map

Abstract The gasoline controlled autoignition (GCAI) is a modern combustion method with which the fuel consumption and the pollutant emissions can be reduced. The major drawback of the combustion method is the limited operating map. In this contribution it is shown how the operating map can be extended towards lower loads by the use of a spark plug for a spark-assisted GCAI combustion. Compared to the GCAI combustion, the spark plug is used additionally and the controller has to be adapted, such that the spark-assisted GCAI combustion is also considered. As controller a model-based predictive controller (MPC) is developed. In this contribution a special focus is set on the investigation of the underlying model for the MPC.

Fuel-Efficient Model-Based Optimal MIMO Control for PCCI Engines

Abstract Recent research in modern combustion technologies, like partial homogeneous charge compression ignition (PCCI), demonstrates the capability of reducing pollutant emissions, e.g. soot and NOX. In addition to this advantage, a possibility to reduce fuel consumption and noise production by model-based optimal control is presented in this paper. In order to understand the basic properties of the PCCI mode, process measurements were conducted using a slightly modified series diesel engine. Control variables are engine combustion parameters: the indicated mean effective pressure, the combustion average and the maximum gradient of the cylinder-pressure. Control inputs are the parameters: quantity of injected fuel, start of injection and the intake manifold fraction of recirculated exhaust gas. The process has very fast, almost proportional behaviour over the engines working cycles. Focusing on the static behaviour of the process, a nonlinear neural network model is used for identification. Successive linearization of the nonlinear network is used to build an affine internal controller model for the actual operating point. The presented controller structure is able to consider constraints by individual formulation of the cost function. With this configuration the closed-loop process is able to track the combustion setpoints with high control quality with minimal possible fuel consumption and combustion noise.

Model Based Control of Gasoline Controlled Auto Ignition

Innovative low-temperature combustion modes for internal combustion engines, such as gasoline-controlled auto-ignition, impose very high requirements on the process control. On one hand, fast reference tracking for the engine load and the combustion phasing is needed, while at the same time, numerous disturbances acting on the highly sensitive process have to be rejected in order to guarantee stable operation at a wide operating range. Model-based predictive control concepts have a great potential to fulfill these requirements. In this contribution, a model-based predictive control consisting of a stationary and dynamic optimization stage is introduced. It is able to account for the characteristic cycle-to-cycle dynamics which occur in gasoline-controlled auto-ignition and also handle constraints imposed on the manipulated and controlled variables of the process.

Actuation Studies for Active Control of Mild Combustion for Gas Turbine Application

MILD combustion (Moderate or Intense Low Oxygen Dilution) is a well known technique that can substantially reduce high temperature regions in burners and thereby reduce thermal NOx emissions. This technology has been successfully applied to conventional furnace systems and seems to be an auspicious concept for reducing NOx and CO emissions in stationary gas turbines. To achieve a flameless combustion regime, fast mixing of recirculated burnt gases with fresh air and fuel in the combustion chamber is needed. In the present study, the combustor concept is based on the reverse flow configuration with two concentrically arranged nozzles for fuel and air injections. The present work deals with the active control of MILD combustion for gas turbine applications. For this purpose, a new concept of air flow rate pulsation is introduced. The pulsating unit offers the possibility to vary the inlet pressure conditions with a high degree of freedom: amplitude, frequency and waveform. The influence of air flow pulsation on MILD combustion is analyzed in terms of NOx and CO emissions. Results under atmospheric pressure show a drastic decrease of NOx emissions, up to 55%, when the pulsating unit is active. CO emissions are maintained at a very low level so that flame extinction is not observed. To get more insights into the effects of pulsation on combustion characteristics, velocity fields in cold flow conditions are investigated. Results show a large radial transfer of flow when pulsation is activated, hence enhancing the mixing process. The flame behavior is analyzed by using OH* chemiluminescence. Images show a larger distributed reaction region over the combustion chamber for pulsation conditions, confirming the hypothesis of a better mixing between fresh and burnt gases.Copyright

A 2-stage MPC Approach for the Cycle-To-Cycle Dynamics of GCAI (Gasoline Controlled Auto Ignition)

Abstract Modern combustion methods, like Gasoline Controlled Autoignition (GCAI), impose very high requirements on the process control. In the investigated set-up, fast reference tracking is needed, while still being able to reject disturbances and satisfy constraints. Model-based predictive controllers (MPC) have a great potential in terms of fulfilling these requirements. In this contribution, a 2-stage MPC controller is introduced. This controller can be used to handle the cycle-to-cycle dynamics of the GCAI process.

Employing the Principal Hessian Direction for Building Hinging Hyperplane Models

In this paper we address the problem of identifying a continuous nonlinear model from a set of discrete observations. The goal is to build a compact and accurate model of an underlying process, which is interpretable by the user, and can be also used for prediction purposes. Hinging hyper plane models are well suited to represent continuous piecewise linear models, but the hinge finding algorithm is guaranteed to converge only in local optima, and hence heavily depends on the initialization. We employ the principal Hessian direction to incorporate the geometrical information of the regression surface in the hinge finding process and can thus avoid the several random initializations proposed in the literature.