Search Results

Abatement and control of emissions from passenger car combustion engines have been in the focus for a long time. Nevertheless, to address upcoming real-world driving emission targets, knowledge of current engine emissions is crucial. Still, adequate sensors for transient emissions are seldom available in production engines. One way to target this issue is by applying virtual sensors which utilize available sensor information in an engine control unit (ECU) and provide estimates of the not measured emissions. For real-world application it is important that the virtual sensor has low complexity and works under varying conditions. Naturally, the choice of suitable inputs from all available candidates will have a strong impact on these factors. In this work a method to set up virtual sensors by means of design of experiments (DOE) and iterative identification of polynomial models is augmented with a novel input candidate selection strategy.

At the moment, no equipment is available for fast measurements of particulate matter (PM) from production CI engines, especially during transients. Against this background, virtual sensors may be an option, provided their precision can be validated. This paper presents a new approach to estimate PM emission based only on in-cylinder pressure data. To this end, an in-cylinder pressure trace is measured with a high resolution (0.5 CAD) and every trace is divided into 8 segments according to critical cylinder events (e.g. opening of the valves or the beginning of injection). A piecewise principle component analysis (PCA) is used to compress the information. This information is then used for PM estimation via a second order polynomial model structure. The key element is the separate use of pressure trace information before and during the early stages of combustion. The model is parameterized by steady points and transient experiments which include parts of the FTP and the NEDC.

The focus in the development of modern exhaust after treatment systems, like the Diesel Oxidation Catalyst (DOC), the Diesel Particulate Filter (DPF) and Selective Catalytic Reduction (SCR), is to increase on one hand the oxidation rates of Carbon monoxide (CO), HC (Hydro Carbons) and NO (Nitrogen Oxide) and on the other hand the reduction rates of Particulate Matter (PM) and the NOx emissions to fulfill the more and more restricting requirements of the exhaust emission legislation. The simplest, practical most relevant way to obtain such a dosing strategy of a SCR system is the use of a nonlinear map, which has to be determined by extensive calibration efforts. This feedforward action has the advantage of not requiring a downstream NOx sensor and can achieve high conversion efficiency under steady-state operating conditions for nominal systems.

The emissions of modern Diesel engines, which are known to have various health effects, are beside the drivers torque demands and low fuel consumptions one of the most challenging issues for combustion and after treatment control. To comply with legal requirements, emission control for heavy duty engines is not feasible without additional hardware, usually consisting of a Diesel oxidation catalyst (DOC), a Diesel particulate filter (DPF) and a selective catalytic reduction (SCR) system. In contrast to other NOx reduction systems, e.g. lean NOx traps, the SCR system requires an additional ingredient, namely ammonia (NH3), to reduce the NOx emissions to non harmful components. Consequently, the correct amount of NH3 dosing in the SCR catalyst is one of the critical components to reach high conversion rates and avoid ammonia slip.

Modern Diesel engines have become complex systems with a high number of available sensor information and degrees of freedom in control. Due to recent developments in production type in-cylinder pressure sensors, there is again an upcoming interest for in-cylinder pressure based applications. Besides the standard approaches, like to use it for closed loop combustion control, also estimation and on-board diagnostics have become important topics. Not surprising in general the trend is to utilize those sensors for as many tasks as possible. Consequently this work focuses on the estimation of the injection parameters based on the indicated pressure signal information which can be seen as first step of a combustion control based on desirable indicated pressure characteristics which may be utilized for e.g. the minimization of NOx emissions. Currently the acquisition of the cylinder pressure traces can be done in real-time by fast FPGA (Field Programmable Gate Array) based systems.

Although SCR is a well established technology for many applications, it is still a field in which several new approaches and components are being tested. Control is a critical issue, as the conflicting requirements of NOx abatement and very small NH3 slip need to be met. Besides empirical solutions, model based controls have been proposed and are probably the technology of choice, also in view of the combination with monitoring functions. However, SCR models are typically based on First Principles (FP), i.e. on global chemical equations and reaction rate equations, and require precise calibration. Still, their performance for the control of dynamic processes is limited, or a high detail, much a priori information, e.g. on the actual SCR reaction rates, are needed. Frequently, this information is not available or reliable, and this is particularly true when components are changed or modified during the development process, so that typically a re-design is needed.

In a typical diesel engine exhaust aftertreatment system consisting of a diesel oxidation catalyst (DOC), a diesel particulate filter (DPF) and a selective catalytic reduction (SCR) system the main purpose of the DOC, besides the oxidation of CO to CO₂, is the oxidation of NO to NO₂. The NO to NO₂ conversion is an essential contribution for the downstream SCR system because the fast SCR reaction which provides the highest conversion rates of NOx to H₂O and N₂ works well only under roughly equal concentrations of NO and NO₂. The typical amount of NO to NOx ratio produced by the engine is about 0.95, hence the DOC is necessary to decrease this coefficient close to 0.50. Due to the temperature dependency of the DOC reaction mechanism the oxidation of NO to NO₂ takes only place sufficiently if the temperature of the DOC is higher than 200°C, which, however, cannot be reached during low engine speed and low load situations.

Pollutant immissions must be kept below some threshold values to prevent health and environmental damage. At the moment, the problem is usually met by constant emission limits for each vehicle independently from specific conditions - in particular, without any relation to the actual immission situation. This approach offers the advantage of simplicity, but offers no guarantee that the immission levels will be kept. New developments, in particular the expected diffusion of i2v methods, allows suggesting context specific emission levels so that the total emission roughly corresponding to the local immissions - can be limited to the target values. To meet this goal, emission-oriented control will be needed. This paper proposes a robust control system which allows tracking a time-varying NOx profile, based on the sliding mode concept.

NOx and PM are the critical emissions to meet the legislation limits for diesel engines. Often a value for these emissions is needed online for on-board diagnostics, engine control, exhaust aftertreatment control, model-based controller design or model-in-the-loop simulations. Besides the obvious method of measuring these emissions, a sensible alternative is to estimate them with virtual sensors. A lot of literature can be found presenting different modeling approaches for NOx emissions. Some are very close to the physics and the chemical reactions taking place inside the combustion chamber, others are only given by adapting general functions to measurement data. Hence, generally speaking, there is not a certain method which is seen as the solution for modeling emissions. Finding the best model approach is not straightforward and depends on the model application, the available measurement channels and the available data set for calibration.