Study of the response of a single-dof dynamic system under stationary non-Gaussian random loads aimed at fatigue life assessment.

Fatigue assessment of components subjected to random loads is a challenging task both due to the variability in amplitude and frequency of the loads and for the computational times required to perform classical time domain fatigue analysis. The frequency domain approach to fatigue life assessment offers a solution by utilizing the power spectral density of the random load, requiring minimal computational effort. However, frequency domain methods are limited to stationary Gaussian signals, while real-world loads often exhibit non-Gaussian characteristics. Previous research proposed formulas to extend frequency domain methods to non-Gaussian cases, but they require knowledge of the parameters related to non-Gaussianity of the component's stress (skewness and kurtosis), which would require a time domain analysis of the stress history on the component and a strong reduction of the computational advantages. This paper aims to address this gap by conducting an extensive campaign of numerical simulations to evaluate the influence of various parameters on the degree of non-Gaussianity of the response of a system. A single-dof mass-spring-damper system was subjected to non-Gaussian random loads of different natures, and the response is analyzed to determine the values of skewness and kurtosis. The study investigated the influence on non-normality indexes of the system's output of several input parameters, which include both the characteristics of the input load and the properties of the dynamic system. The findings contribute to a better understanding of non-Gaussianity in dynamic systems and pave the way for conducting efficient fatigue analyses in the frequency domain. Future work will extend the study to non-stationary random loads, further advancing the understanding of non-Gaussianity and non-stationarity in dynamic systems.

Deep learning models to predict flood events in fast-flowing watersheds.

This study aims to explore the reliability of flood warning forecasts based on deep learning models, in particular Long-Short Term Memory (LSTM) architecture. We also wish to verify the applicability of flood event predictions for a river with flood events lasting only a few hours, with the aid of hydrometric control stations. This methodology allows for the creation of a system able to identify flood events with acceptable errors within several hours' notice. In terms of errors, the results obtained in this study can be compared to those obtained by using physics-based models for the same study area. These kinds of models use few types of data, unlike physical models that require the estimation of several parameters. However, the deep learning models are data-driven and for this reason they can influence the results obtained. Therefore, we tested the stability of the models by simulating the missing or wrong input data of the model, and this allowed us to achieve excellent results. Indeed, the models were stable even if several data were missing. This method makes it possible to lay the foundations for the future application of these techniques when there is an absence of geological-hydrogeological information preventing physical modeling of the run-off process or in cases of relatively small basins, where the complex system and the unsatisfactory modeling of the phenomenon do not allow a correct application of physical-based models. The forecast of flood events is fundamental for correct and adequate territory management, in particular when significant climatic changes occur. The study area is that of the Arno River (in Tuscany, Italy), which crosses some of the most important cities of central Italy, in terms of population, cultural heritage, and socio-economic activities.