System identification of inverse, multimodal and nonlinear problems using evolutionary computing – Application to a pile structure supported on nonlinear springs

• System identification of a mechanical structure supported by nonlinear springs.
• Optimization of a nonlinear and inverse problem using evolutionary computing (EC).
• Combination of EC with problem specific a priori knowledge.
• The practical application is focused on the monitoring of axially loaded piles.

This paper deals with the system identification of a mechanical structure supported by nonlinear springs subjected to an external load. If all mechanical parameters of the system were known, the displacement of the system subjected to this load could be easily calculated. However, the monitoring applications often deal with the inverse problem. The loads and displacements of the system are known and certain mechanical parameters of the system are sought. The solution of such inverse problems can be difficult, especially when they have a nonlinear and multimodal character, which often makes them appear intractable at first sight.However, evolutionary computing can be applied to solve this inverse, nonlinear and multimodal problem. Sometimes a prior knowledge exists on certain system properties, which is difficult to implement into analytical or numerical solvers. This knowledge can play a decisive role in identifying the system properties and it can be easily included as a boundary condition when applying evolutionary algorithms. This article discusses how and under what conditions the unknown spring resistances can be identified. The practical application of this procedure is exemplified here with the mechanical system of a pile foundation.