In Elastic Optical Networks (EON), deep learning can be used to facilitate automated processes to unlock higher optical layer capability. Rerouting without making a distinction between an intense breakdown and an easier breakdown, however, could create an issue with low network capacity utilization. A deep learning-based detecting the link-failure method is suggested to identify and customize the breakdown in EON in order to solve this issue. Networks like the Internet of Things (IoT) are growing exponentially and being more complicated with diverse wireless and wired connections as a consequence of the rapid advancements in technological innovation and the increasing number of network components. In these systems, link failures might cause a link to be disconnected without an immediate substitute or reconnected, as when a wireless module switches to a different access point. Localizing the failing connection and determining whether the link becomes separated or reconnected has become difficult problems. An active testing strategy incurs significant latency for communication and expense since it takes many hours to investigate the system via signals on several pathways. In this paper, optimal routing with link failure detection model is developed to effectively detect the malicious nodes and faults in networks for providing efficient transmission in EONs. First, the link failure detection is performed using Long Short-Term Memory (LSTM) to detect the link failures. Then, the optimal routing is performed using Whale Optimization Algorithm (WOA) algorithm for effectively selecting the optimal path. Multi objective constraints analysis like shortest path, link reliability and stability, throughput are enhanced by the efficiency of the developed model. The developed link failure detection model performance is compared to various heuristic algorithms and it showed high accuracy.