Fredrik Östman

Adaptive Cylinder Balancing of Internal Combustion Engines

For internal combustion engines with electronic fuel-injection systems, differences in fuel injector performance may result in significant discrepancies between the fuel amounts injected into the individual cylinders. In order to avoid excessive torsional vibrations of the crankshaft due to nonuniform cylinder-wise torque contributions, it is important to calibrate the fuel injectors. This can be done using angular speed measurements of the crankshaft to reconstruct the gas torque applied on the flywheel and relating it to the cylinder firings. For engines with a high number of cylinders, an accurate engine model is essential for successful cylinder balancing. This is due to closely spaced cylinder firings and the fact that the crankshaft dynamics cannot be ignored, partly due to the increased length of the crankshaft, and partly because analysis of higher frequency components is required to obtain sufficient information for balancing the cylinder-wise torque contributions. The need for accurate models forms a major obstacle to the implementation of cylinder balancing methods for such engines. In this paper an online adaptive cylinder balancing method is proposed which can be applied to engines with unknown crankshaft dynamics. The procedure identifies the engine-specific phase-angle diagrams which relate the amplitudes and phases of the relevant torque frequencies to the cylinder firings. The identified model is applied to balance the cylinder-wise torque contributions of the engine with the objective of minimizing torsional vibrations. The proposed method is evaluated using both simulations on twelve and eighteen cylinder V-engine generator sets and experiments on a full-scale 6 MW six cylinder common-rail diesel power plant engine. The results show that the adaptive cylinder balancing procedure achieves a more efficient reduction of torsional vibrations than methods based on a fixed engine model, in which a rigid crankshaft is assumed.

Engine Knock Margin Estimation Using In-Cylinder Pressure Measurements

Engine knock is among the most relevant limiting factors in the improvement of the operation of spark-ignited engines. Due to an abnormal combustion inside the cylinder chamber, it can cause performance worsening or even serious mechanical damage. Being the result of complex local chemical phenomena, knock turns out to have a significant random behavior but the increasing availability of new on-board sensors permits a deeper understanding of its mechanism. The aim of this paper is to exploit in-cylinder pressure sensors to derive a knock estimator, based on the logistic regression technique. Thanks to the proposed approach, it is possible to explicitly deal with knock random variability and to define the so-called margin (or distance) from the knocking condition, which has been recently proven to be an effective concept for innovative knock control strategies. In a model-based estimation fashion, two modeling approaches are compared: one relies on well-known physical mechanisms while the second exploits a principal component analysis to extract relevant pressure information, thus reducing the identification effort and improving the estimation performance.

Engine parameter outlier detection: verification by simulating PID controllers generated by genetic algorithm

We propose a method for engine configuration diagnostics based on clustering of engine parameters. The method is tested using simulation of PID controller parameters generated and selected using a genetic algorithm. The parameter analysis is based on a state-of-the art method using multivariate extreme value statistics for outlier detection. This method is modified using a variational mixture model which automatically defines a number of Gaussian kernels and replaces a Gaussian mixture model.

Power Balancing of Internal Combustion Engines – A Time and Frequency Domain Analysis

Abstract Power balancing of internal combustion engines is studied in time and frequency domains. Speed measurements at the flywheel and the load side are used to estimated the torque applied to the flywheel. The power contributions during the cylinder-wise work phases are estimated by integrating the reconstructed torque, and a controller which adjusts the fuel injections to reduce power unbalances is applied. A convergent power balancing algorithm is presented for multi-cylinder engines, where the work phases of the individual cylinders may overlap. This time-domain method is verified against previous frequency-domain approaches. The analysis shows that for engines with flexible crankshafts, the time-domain method performs in general differently from frequency-domain cylinder balancing approaches. The feasibility of the power balancing algorithm is investigated by simulations and full-scale tests on a large diesel power plant engine.

Regulator Design using a Multiobjective Genetic Algorithm

Abstract When designing a regulator for a specific system, a set of different objectives needs to be evaluated. This is usually done by manual tuning of parameters and evaluation of the step response. In recent years gradient-based multiobjective constraint optimization methods have been successfully applied on this problem while, in this report, a multiobjective genetic algorithm approach is applied. By defining a set of cost-functions in the frequency-domain for different frequency intervals, PID regulator settings for two bench-mark models are identified and successfully evaluated.