Vehicle Driveability Assessment using Neural Networks for Development, Calibration and Quality Tests

2000-01-0702

03/06/2000

Event
SAE 2000 World Congress
Authors Abstract
Content
Actual automotive themes in the beginning century are globalization and platform concepts. Platforms reduce manpower for basic power train development and enable a higher vehicle quality by sharing development cost to many models.
New drive train generations with direct injected diesel and gasoline engines, variable valve train systems and hybrid drives require complex electronic control systems with many control parameters, which must be calibrated for each platform model to fulfill the targets for emissions, diagnostics and driveability. Calibration becomes a critical procedure in vehicle development. A negative effect of the platform is the reduced possibility to give a model or an OEM a brand specific driveability character, traditionally an important sales - promoting factor.
The paper describes a tool for the objective real time assessment of vehicle driveability and vehicle character, using a new subjective - objective approach. Driveability is described by a high number of 275 single criteria, including vehicle dynamics, noise and vibration. For each criteria an objective rating is calculated on line during driving, including averages for driving modes and criteria. Ratings and averages are derived from experienced company test drivers. Vehicle class expectation is considered by the definition of seven vehicle classes for AT and MT applications.
The tool, representing an artificial driver, can be used for several vehicle development activities, e.g. ECU calibration, benchmarking and quality tests. Special interfaces for vehicle, chassis dyno and highly dynamic engine test bed enable applications during all development phases.
The advantages are a more than 10 times faster duration of driveability analysis, fully reproducibility, and common description for OEM, supplier and customer, engine and vehicle development engineers. The prediction of the vehicle driveability quality on test bed enables an early detection of later on driveability problems and enables more precise adjustment. The simultaneous measurement of performance, emissions, fuel economy and driveability leads to better overall development results.
Efficient calibration can be achieved manually or computer assisted by using tools for “Design of Experience” and “Automatic calibration”. A new driveability “brand” design can be created for vehicles. Higher driveability quality increases the customer satisfaction and leads to an increased driving pleasure. In addition a new driveability “brand” design can be created for vehicles.
Meta TagsDetails
DOI
https://doi.org/10.4271/2000-01-0702
Pages
12
Citation
Schoeggl, P., and Ramschak, E., "Vehicle Driveability Assessment using Neural Networks for Development, Calibration and Quality Tests," SAE Technical Paper 2000-01-0702, 2000, https://doi.org/10.4271/2000-01-0702.
Additional Details
Publisher
Published
Mar 6, 2000
Product Code
2000-01-0702
Content Type
Technical Paper
Language
English