The objectives of Bosch in automotive applications are to develop and improve top-down design methods for analogue and mixed analogue/digital ICs. This paper shows the investigation of a method which is based on a neuro-fuzzy approach. The idea is to automatically generate simplified models describing the non-linear dynamic behaviour of analogue blocks using neurofuzzy identification methods. These simplified models are required to reduce the simulation time when validating whole complex systems. The chosen neuro-fuzzy identification method has been realized by using linear filters represented in the form of a state space description. This method automatically selects the required non-linear dynamic models and also determines the parameters of the model algorithms fully automatically. Afterwards these models are automatically converted into block diagram descriptions for Simulink. The necessary training data are generated within Simulink using Saber through the co-simulation link InSiMS. Based on automotive electronic examples it was shown that the use of dynamic neuro-fuzzy models results in a reduction in simulation time at least by a factor of 20 compared to a Saber simulation. The accuracy of the simulation results was always in the given tolerance range. A significant advantage of this approach and its application to mixed-signal systems is its computational efficiency which results in a significant reduction of simulation time. It is shown that the proposed method is capable of simulating SISO and MMO systems with high accuracy and has in addition the potential to interpret unseen data with acceptable results.
Automatic identification method for analog and mixed analog/digital ICs in automotive electronics
Automatisches Identifikationsverfahren für Analog- und Mixed-Signal-ICs in der Automobilelektronik
2004
10 Seiten, 9 Bilder, 9 Quellen
Aufsatz (Konferenz)
Englisch
analoge integrierte Schaltung , digitale integrierte Schaltung , dynamisches Verhalten , Fuzzy-Logik , Identifikation , integrierte Halbleiterschaltung , Kraftfahrzeugelektronik , Merkmalerkennung , Modellsimulation , neuronales Netzwerk , Simulationsmodell , Simulationsmodellbildung , Systemidentifikation , Zulieferbetrieb
Mixed Analog-Digital WDM Local Area Networks
IEEE | 2006
|An Analog/Digital Integrated Circuit Interface for Automotive Sensors
SAE Technical Papers | 1976
|An analog/digital integrated ciruit interface for automotive sensors
Kraftfahrwesen | 1976
|Mixed analog/digital silicon compiler for space applications
Tema Archiv | 1993
|