Search Results

Vehicle development involves the design and integration of subsystems of different domains to meet performance, efficiency, and emissions targets set during the initial developmental stages. Before a physical prototype of a vehicle or vehicle powertrain is tested, engineers build and test virtual prototypes of the design(s) on multiple stages throughout the development cycle. In addition, controllers and physical prototypes of subsystems are tested under simulated signals before a physical prototype of the vehicle is available. Different departments within an automotive company tend to use different modelling and simulation tools specific to the needs of their specific engineering discipline. While this makes sense considering the development of the said system, subsystem, or component, modern holistic vehicle engineering requires the constituent parts to operate in synergy with one-another in order to ensure vehicle-level optimal performance.

A recent trend in powertrain development organisations has been to apply processes historically associated with manufacturing. The aim is to capitalise on the resulting productivity gains to contain the increasing test demand necessary to develop current and future product. Significant obstacles to the implementation of manufacturing derived methods include the lack of clarity of the engineering test requirements and existing working practices in the product development environment. The System Optimisation Approach has been presented in previous work as a potential solution [1]. As an extension, this paper introduces a new concept closely related to the established manufacturing principle of Process Capability (Cp). Application of the resulting method quantifies the test facility’s capability to provide a test result subject to a specified statistical confidence within a certain number of test repeats.

This paper documents some of the findings from a joint JLR and AVL project which was conducted at the JLR Gaydon test facility in the UK. A testing and development efficiency concept is presented and test data quality is identified as a key factor. In support of this methods are proposed to correctly measure and set targets for data quality with high confidence. An illustrative example is presented involving a Diesel passenger car calibration process which requires response surface models (RSMs) of key engine measured quantities e.g. engine-out emissions and fuel consumption. Methods are proposed that attempt to quantify the relationships between RSM statistical model quality metrics, test data variability measures and design of experiment (DOE) formulation. The methods are tested using simulated and real test data.