Energy-Efficient Fault Data Prediction and Transmission in WSN IoT using Bio-Inspired Optimization and Deep Learning

Wireless sensor networks (WSNs) are crucial for several applications. WSN nodes frequently function with constrained battery capacity, rendering energy efficiency a critical issue for clustering and routing. Moreover, a principal challenge of WSNs is ensuring the dependability and security of transmitted data in susceptible contexts to avert hostile node attacks. This study seeks to establish a secure and energy-efficient routing system for fault data prediction to improve the longevity and dependability of WSNs. This paper presents a sophisticated framework for intelligent fault prediction and energy-efficient data transmission in WSN, utilising bio-inspired optimisation and deep learning methodologies. The model initiates data fault prediction with Multi-Term Fourier Graph Neural Networks (MTFGNN), which examine temporal and spatial relationships to detect anomalies and defective nodes prior to clustering. Faultless nodes are subsequently categorised by Fuzzy C-Means (FCM) clustering, facilitating adaptive and efficient cluster creation. Quokka Swarm Optimisation (QSO) is utilised to improve energy efficiency by selecting ideal cluster heads (CH), thereby balancing energy usage and reducing intra-cluster communication expenses. A trust-based routing technique employs Proximal Policy Optimisation (PPO), a reinforcement learning method that dynamically identifies secure and energy-efficient pathways for data transfer, while reducing the influence of unreliable nodes. The experimental results indicate that it surpasses the rival methods across multiple performance parameters. The performance outcomes of quality of service (QoS) metrics are delineated as follows: energy consumption (0.204), throughput (0.701), packet delivery rate (94.24%), network lifetime (1310 rounds), and fault prediction accuracy (99.78%), precision (98.69%), recall (97.52%) and F1 score (97.83).