Search

The research described in this paper was aimed at demonstrating the implementation of statistically derived tolerances to so-called ‘noise’ factors that cause imprecision in the vehicle fuel consumption during chassis dynamometer testing. These tolerances were derived from previous work carried out by one of the authors co-workers and were set to achieve a repeatability target of 0.5% coefficient of variation in the vehicle fuel consumption. This target was successfully achieved during a test programme to determine the fuel consumption benefit of two candidate engine oils over production engine oil using a 1.0 l gasoline passenger test vehicle. Regression response modelling was used to determine whether the recorded variability was correlated with the variability in the vehicle fuel consumption and it was found that all the measured test noise factors were adequately controlled. A universal methodology is proposed for the use of the response modelling technique to verify adequate control of known noise factors and to allow for corrections to the vehicle fuel consumption to be performed where factors have not been adequately controlled, without the need to complete additional testing.

The paper details improvements in the technique used to locate the boundary of the permissible operating envelope (or limit space) of an engine. This process is an important step in the model-based calibration of engines and is currently conducted manually or using cautious automatic search algorithms because of the risks to engine integrity if operating limits are exceeded. To improve this process, prior knowledge was used to mimic the behaviour of an expert calibrator following two strategies. The first was designed to offer a rapid search based on a similar premise to commercially available tools. The second scheme offers a thorough search of the limit space using a scheme that targets the extremes of the operating envelope. Both patterns were enhanced using prior knowledge generated from previous operating points in order to speed up the search. When implemented on the test bed the use of prior knowledge in this way has shown the potential to shorten the search time by up to 25 per cent. The thorough search pattern increased the accuracy of the limit space definition by up to 40 per cent but resulted in a significantly slower search.

Cyclic variability (CV) in combustion has been studied over many years and is still an active area of research. Despite the application of many analytical tools, including, in more recent years, those connected with the study of non-linear dynamic systems, CV remains a challenging subject. In the past, most researchers into CV have presented their findings in the form of overall statistical measures. However, it is felt that perhaps more attention needs to be paid to sequences of combustion cycles on a detailed cycle-to-cycle basis, particularly where the aim is to try to formulate a control mechanism to reduce CV, and that sonification combined with self-similarity plots provides a useful tool to facilitate such an approach. The purpose of this paper is to introduce sonification to automotive engineers, who may be unfamiliar with it, and to give an example of its implementation and application in the study of CV in internal combustion engines.

This paper describes the development of an exhaust system model based on the characterization of heat transfer in a series of 11 test sections designed to represent a range of configurations seen in production exhaust systems. These sections include variations in the wall thickness, diameter, bend angle, and radius. This work is part of a larger activity aimed at the accurate modelling of heat transfer and subsequent catalyst light-off in production exhaust systems consisting of similar geometries. For each section a range of steady state tests was performed on a dynamic test cell using a port injected gasoline engine. In each case a correlation between the observed Reynolds number Re and the Nusselt number Nu was developed. A model of the system was implemented in MATLAB—Simulink in which each pipe element was split into 25 subelements by dividing the pipe into five, both axially and radially. After each individual section had been characterized, a generic model was built that was capable of determining the relationship between the Nusselt number and the Reynolds number for previously unseen pipe geometries. Complex exhaust systems can be constructed from smaller elements of defined geometry. The modelling approach was validated using experimental data gathered from vehicle testing on a chassis dynamometer. The steady state relationship between Re and Nu allows heat transfer in the test section to be predicted over transient test cycles and demonstrates good agreement with relationships in the literature.