Computational Risk Modeling In Engineering Systems Under Uncertainty For Concrete Gravity Dam Safety

Abstract

In engineering systems, risk modeling studies material uncertainty, loading variation, and operational disturbances. Despite the advancements in risk modeling, there have been deficiencies concerning the incorporation of adaptive intelligence and uncertainty-aware learning in providing appropriate decisions on complex earthquakes. This research seeks to design an intelligent model for computational risk modeling for assessing the safety of concrete gravity dams in case of uncertainty through EFO-ISVM. In this regard, a detailed dataset with 10,000 cases and 49 features such as dam height, reservoir height, earthquake strength, foundation quality, concrete strength, and risk level is employed. For pre-processing, Min-Max normalization technique is used while PCA is considered for the feature extraction process. Furthermore, the Electric Fish Optimizer (EFO), which uses electric-field-based movements to tune hyperparameters effectively, is utilized here. Intelligent Support Vector Machine (ISVM) implements nonlinear classification and regression to compute risk levels and failure probability in response to uncertain data. Risk modeling is done using probabilistic inference and scenario analysis. The results of the experiments illustrate the high efficiency of risk prediction by the proposed algorithm with accuracy of 0.920, sensitivity of 0.930, specificity of 0.970, and Area Under the Curve of 0.940. This model was coded in Python 3.11 with the help of Jupyter Notebook and Scikit-learn, NumPy, Pandas, SciPy, and Matplotlib libraries. The suggested system is helpful for risk prioritization, thus supporting more effective decision-making for assessing safety levels of concrete gravity dams.