Towards a computational engineering tool for structural sensitivity analysis based on the method of influence functions

This contribution presents a computational engineering tool for the purposeful and systematic investigation how sensitive responses of a structural analysis model react due to variations of model parameters. The methodical basis of the approach is the relationship between adjoint sensitivity analysis and the classical and well-known technique of influence functions. The strong connection between the two approaches allows the extension of the traditional method of influence functions as a tool for the evaluation of sensitivities for general responses, parameters and structures. The paper provides a detailed discussion of that methodical relationship and shows how the influence function of important responses in structural engineering can be identified in the adjoint sensitivity equation. Following the ability of the traditional method of influence functions to provide structural comprehension as visual means, also the results of sensitivity analysis are graphically processed as maps. Finally, the application of the presented method and the sensitivity maps is demonstrated by means of expressive engineering structures. The examples provide insights how sensitivity analysis can support the structural design process by fast and purposeful identification and further assessment of the importance of model parameters.