Establishing criteria to ensure successful feedforward artificial neural network modelling of mechanical systems

The emulation of mechanical systems is a popular application of artificial neural networks in engineering. This paper examines general principles of modelling mechanical systems by feedforward artificial neural networks (FFANNs). The slow convergence issue associated with the highly parallel and redundant structure of FFANN systems is addressed by formulating criteria for constraining network parameters so that FFANNs may be reliably applied to mechanics problems. The existence of the FFANN mechanical model and its stability during construction, with respect to the error in the data, are analyzed. Also, a class of differential equations is analyzed for use with Tikhonov regularization. It is shown that the use of Tikhonov regularization can aid in FFANN data-driven construction with a priori mathematical models of varying degrees of physical fidelity. Criteria to ensure successful FFANN application from an engineering perspective are established.