Abstract-This paper presents two torque estimation methods for vehicle engines: unknown input observer (UIO) and adaptive parameter estimation. We first propose a novel yet simple unknown input observer based on the crankshaft rotation dynamics only. For this purpose, an invariant manifold is derived by defining auxiliary variables in terms of first order low-pass filters, where only one constant (filter coefficient) needs to be tuned. These filtered variables are used to calculate the estimated torque. Robustness of this UIO against sensor noise is studied and compared to two other estimators. On the other hand, since the engine torque dynamics can be formulated as a parameterized form with unknown time-varying parameters, we further present several adaptive laws for time-varying parameter estimation. The parameter estimation errors are derived to drive these adaptive laws and time-varying adaptive gains are introduced. The two proposed estimators only use the measured air mass flow rate and engine speed, and thus allow for improved computational efficiency. Both estimators are verified via a dynamic engine simulator built in a commercial software GT-Power (Ricardo Wave), and also practically tested via experimental data collected in a dynamometer test-rig. Both simulations and practical results show very encouraging results with small estimation errors even in the presence of sensor noise.Index Terms-Engine torque estimation, mean value engine model, unknown input observer, time-varying parameter estimation.

This brief addresses the emission reduction of spark ignition engines by proposing a new control to regulate the air-fuel ratio (AFR) around the ideal value. After revisiting the engine dynamics, the AFR regulation is represented as a tracking control of the injected fuel amount. This allows to take the fuel film dynamics into consideration and simplify the control design. The lumped unknown engine dynamics in the new formulation are online estimated by suggesting a new effective unknown system dynamics estimator. The estimated variable can be superimposed on a commercially configured, wellcalibrated gain scheduling like proportional-integral-differential (PID) control to achieve a better AFR response. The salient feature of this proposed control scheme lies in its simplicity and the small number of required measurements, that is, only the air mass flow rate, the pressure and temperature in the intake manifold, and the measured AFR value are used. Practical experiments on a Tata Motors Limited two-cylinder gasoline engine are carried out under a realistic driving cycle. The comparative results show that the proposed control can achieve an improved AFR control response and reduced emissions. Index Terms-Air-fuel ratio (AFR) control, lambda sensor, spark ignition (SI) engines, unknown dynamics estimator. I. INTRODUCTION T HE requirement for engine emissions has become more stringent in recent years. To reduce emissions, spark ignition (SI) engines are usually configured with a three-way catalyst (TWC) to convert the pollutant exhaust into innocuous gases [1]. However, it is of great importance that the airfuel ratio (AFR) in the combustion chamber is maintained at the ideal value because the catalyst conversion efficiency,

Abstract-This paper presents a novel adaptive controller of air-fuel ratio (AFR) in spark ignition (SI) engines. The controller robustly estimates unknown time-varying engine parameters and thus improves both the transient and steady-state performance. The objective is to regulate the AFR in the combustion chamber around the stoichiometric value by manipulating the injected fuel mass flow rate so as to improve fuel economy and to reduce emissions. The AFR regulation problem is first reformulated into a tracking control problem of the fuel mass flow. This simplifies the control synthesis, i.e. the number of parameters to be online updated can be reduced. Then a representation of the parameter estimation error is derived by using auxiliary filter operations, which is used as a new leakage term in the adaptive law. In this case, exponential convergence of the AFR error and the estimation of the time-varying parameters can be proved simultaneously. The proposed controller is compared with a generic adaptive controller using the gradient descent method based on a well-calibrated mean value engine model (MVEM). Finally, the proposed controller is also validated with a commercial engine simulation software, GT-Power, demonstrating better results for the suggested adaptive controller than for the gradient descent approach.

The use of Wankel engines has been severely limited as the emission regulations get stringent around the world since the 1970s. The fuel puddles due to port fuel injection (PFI) and the leakage between combustion chambers are significant sources of efficiency loss and emissions. For most spark ignition engines in production, the emission strongly depends on the air-fuel ratio (AFR) controller in cooperation with a three-way catalytic (TWC) converter. This paper presents a generic observer-based AFR control framework to deal with the high nonlinearities of Wankel engines so as to improve the fuel economy and emissions. By taking the unknown parameters as augmented engine states, an extended Kalman filter is designed to estimate the fuel puddle dynamics using only mass air flow (MAF) and lambda sensors. The complex nonlinear air-filling dynamics are lumped together and estimated using novel observer techniques. A newly proposed unknown input observer is compared with a dirty differentiation observer and then employed in the feedback AFR control design. Comparative simulations based on a calibrated benchmark engine model show that the proposed control can speed up the transient response and regulate the AFR around the stoichiometric value.

T he use of Wankel rotary engines as a range extender has been recognised as an appealing method to enhance the performance of Hybrid Electric Vehicles (HEV). They are effective alternatives to conventional reciprocating piston engines due to their considerable merits such as lightness, compactness, and higher power-to-weight ratio. However, further improvements on Wankel engines in terms of fuel economy and emissions are still needed. The objective of this work is to investigate the engine modelling methodology that is particularly suitable for the theoretical studies on Wankel engine dynamics and new control development.In this paper, control-oriented models are developed for a 225CS Wankel rotary engine produced by Advanced Innovative Engineering (AIE) UK Ltd. Through a synthesis approach that involves State Space (SS) principles and the artificial Neural Networks (NN), the Wankel engine models are derived by leveraging both first-principle knowledge and engine test data. We first re-investigate the classical physicsbased Mean Value Engine Model (MVEM). It consists of differential equations mixed with empirical static maps, which are inherently nonlinear and coupled. Therefore, we derive a SS formulation which introduces a compact control-oriented structure with low computational demand. It avoids the cumbersome structure of the MVEM and can further facilitate the advanced modern control design. On the other hand, via black-box system identification techniques, we compare the different NN architectures that are suitable for engine modelling using time-series test data: 1) the Multi-Layer Perceptron (MLP) feedforward network; 2) the Elman recurrent network; 3) the Nonlinear AutoRegressive with eXogenous inputs (NARX) recurrent network. The NN models overall tend to achieve higher accuracy than the MVEM and the SS model and do not require a priori knowledge of the underlying physics of the engine.