Feedforward artificial neural network for off-nominal spark control

US11125202B1 · kind B1 · utility

Assignee

• FCA US LLC. · US

Inventors

• Shuonan Xu · Troy, US
• Ethan A Bayer · Lake Orion, US
• William P Attard · Brighton, US
• David A Lawrence · Lake Orion, US
• Nandanshri Bagadi · Shelby Charter Township, US

Key dates

• Filing date: Jul 15, 2020
• Grant date: Sep 21, 2021
• Expiry date: Jul 15, 2040

Classification

• Technology area (CPC Y): Emerging Cross-Sectional Technologies
• CPC primary: Y02T10/40
• WIPO field: Engines, pumps, turbines
• WIPO sector: Mechanical engineering

Abstract

Engine combustion phasing control techniques utilize a trained feedforward artificial neural network (ANN) to model both base and maximum brake torque (MBT) spark timing based on six inputs: intake and exhaust camshaft positions, mass and temperature of an air charge being provided to each cylinder of the engine, engine speed, engine coolant temperature. The selected target spark timing could be adjusted based on a two-dimensional surface having engine speed and air charge mass as inputs. The target spark timing adjustment could be performed only during an initial period when the trained ANN is immature. The ANN could also be trained using dynamometer data for the engine that is artificially weighted for high load regions where accuracy of spark timing is critical.