Ein Beitrag zur Beherrschung von Unsicherheit in Lastmonitoring-Systemen

A fundamental characteristic for the lifetime of load-carrying structures is the operating load. Real operating loads for an individual structure are mostly only partly available from rough estimations for similar structures. A load-monitoring system provides the identification of current load and deformation states during the usage of load-carrying structures. Normally, it consists of sensors, a mathematical model to represent the relevant systems properties and a suitable information processing algorithm. In this work, an observer-based estimation algorithm for unknown input is used to identify the current loads and deformation states. A new projection approach is presented to solve socalled underdetermined inverse problems. This approach is based on the assumption that often only a few causes result to the observed effects. The approach is used in numerical and experimental simulations for two practical examples in order to a) identify the location of a load and b) identify changing system properties without knowing loads.
Uncertainty that may occur within the processes of load monitoring could potentially lead to a deviation between the real loads and those that are identified by the load-monitoring. This uncertainty will be controlled by the methodology developed within this work. The methodology is based on the usage of known and refined methods to describe, evaluate and eventually control uncertainty. The contribution of each method shown in this work is rated and compared qualitatively and quantitatively in numerous analytical, numerical and experimental investigations. For a qualitative rating, an uncertainty model is used to classify uncertainty based on the level of available information. The root mean squared error of the deviation between directly measured loads and loads that are estimated by the load-monitoring is used for quantitative rating. On the one hand, the highest amount of uncertainty results from the unavoidable model uncertainty due to the simplified and idealized representation of the real structure in a mathematical model. This model uncertainty is statistically described by using a model error model. On the other hand, the choice of sensor position has the highest amount to reduce uncertainty in a load-monitoring system.
The functional efficiency of the proposed load-monitoring system is proven by numerical and experimental simulations using the example of an axially and laterally loaded beam structure and a plate. For a broadband excitation in experimental simulations, a relative root mean squared error between measured and identified load between 4% and 12% occurs.
The present work is a contribution to reduce uncertainty in the usage of load-carrying structures by providing methods to identify individual operation loads. Thereby, uncertainty that may occur during the processes of load-monitoring can be controlled in a holistic and integrated manner for the first time.