Multimodal data optimization for structural health monitoring using a deep Recursive Gaussian fuzzy Variational encoder–Reinforcement Quantile neural network

The use of sensors was crucial in the detection of damage signals such as cracks and vibrations through structural health monitoring (SHM) systems. The application of machine learning (ML) to the data obtained from sensors improved SHM by enabling the detection of damage patterns that would otherwise go unnoticed by human analysts. This research proposes a novel technique in SHM by multimodal data feature selection with optimization using ML algorithms for anomaly detection. The input to the proposed system consists of SHM-based multimodal anomaly data, which are preprocessed through noise removal and normalization. Processed data feature selection and classification are carried out using a Deep Recursive Gaussian Fuzzy Variational Encoder with a Reinforcement Quantile Neural Network (DRGFVE-RQNN). Then, the classified features are optimized using the Genetic Chaotic Binary Whale Swarm Differential Dragonfly Evolution algorithm (GCBWSDDE). The experimental investigation is conducted in terms of processing time, dependability, scalability, detection accuracy, and correlation coefficient (CC). The suggested approach has a bright future as a trustworthy digital tool with AI support for evaluating safety in SHM systems.