Black box modelling of a two-stroke racing motorcycle engine for virtual prototyping appplications

In the design process of complex systems, the use of physical prototypes presents many drawbacks in terms of both costs and time. Virtual prototyping tools allow instead the designer to extensively study the system behavior in a variety of configurations, providing the possibility of performing focussed tests and comparing the effectiveness of different solutions before the physical realization of the prototype. When building the virtual prototype, the choice of a suitable model for each of the interacting components is crucial for achieving reliable results. In the simulation of racing vehicles, such as cars or motorcycles, it is common practice to represent the engine torque by means of stationary maps depending on a finite number of constant values of the throttle fraction and the rotational speed of the crankshaft. Although simple to implement and characterize, such model cannot reproduce dynamic transients, and therefore may not be adequate when a detailed analysis of the vehicle behavior is required. In this paper we present the characterization and identification of a nonlinear dynamic model for a two-stroke internal combustion high-performance engine, to be used in the ADAMS multibody virtual prototyping environment. The model is obtained by using neural networks and validated on experimental data