Vehicle type approval drive cycles have become a mainstay for benchmarking performance of engines in the development cycle. However, they are typically long and costly to test, with questions of repeatability and real-world relevance. With the move to Real Driving Emission (RDE) style testing, complete foreknowledge of the cycle is no longer guaranteed. This paper presents a methodology for identifying key behaviours (or information rich regions) from the current worldwide harmonised light-duty test cycle (WLTC) type approval test using a moving 2-minute window approach. Three techniques for pattern recognition are presented and applied to data collected from a modern Gasoline Turbocharged Direct Injection (GTDI) engine, run through the WLTC. The techniques examine different points in the process, with the first examining response data, and working backwards to the original vehicle speed cycle specification. The first two techniques, Intensity Ratio (IR) for cumulative responses and Energy Residency (ER) for engine inputs, are newly developed in this paper. The final technique, Dynamic Time Warping (DTW) is a new application of an existing tool to the subject of vehicle drive cycles. The techniques are examined in isolation and discussed, before being brought together to identify commonly agreed information rich sub-cycle candidates that best represent the parent cycle. The techniques make use of a combination of time windowing, signal derivative and peak analysis methods. The two-minute window is chosen based on the length of an existing sub-cycle that was identified as part of earlier work, and this method is also described and validated in this paper. One sub-cycle identified by the approach represents a duration reduction of 93% to a containable 2-minute transient. This segment accounts for 15% of the overall WLTC fuel used. A discussion of the techniques and their applications is presented to motivate future work in this area.

This content is only available via PDF.
You do not currently have access to this content.