Machine Learning Based Fault Detection And Recovery Framework For Iot Environmental Monitoring Applications

Abstract

In Internet of Things (IoT) enabled environmental monitoring sector, fault detection and recovery systems are crucial for guaranteeing continuous, accurate, and reliable service delivery. IoT devices often function in resource-constrained environments, making the implementation of complex fault-detection algorithms difficult. In this paper, Automatic Fault Detection and Recovery (AFDR) framework based on Artificial Neural Network Fuzzy Inference System (ANFIS) is proposed. In this framework, the device faults along a path are determined based on the ANFIS model. Fuzzy rules are provided based on the packet loss rate (PLR), Signal to Interference Noise Ratio (SINR) and round trip delay (RTD) metrics. In the fault recovery phase, the recovery agent at the primary path establishes an alternate fault-free route by excluding the faulty nodes. The evaluation metrics packet delivery ratio (PDR), number of packets dropped, average residual energy computational cost and end-to-end delay are measured by varying the number of fog nodes.Experimental results show that the proposed AFDR-ANFIS model attains higher fault detection accuracy with reduced packet drops and computational overhead. The proposed Machine Learning (ML) based technique plays a vital role in fault detection and recovery systems in IoT-based healthcare.