An Adaptive Neural Identifier with Applications to Financial and Welding Systems

This paper considers the online identification problem of uncertain systems. Based on parallel and series-parallel configurations with feedback and by using Lyapunov arguments, a unified identification algorithm is introduced to ensure the boundedness of all associated errors and convergence of the state estimation error to an arbitrary neighborhood of the origin. The main peculiarity of the proposed algorithm lies in allowing the adjustment of the identification transient by using parameters that are not related to the residual state error. Two examples are deemed to validate the theoretical results and show the relevance of the application of the proposed methodology for online weld geometry prediction.