An efficient methodology to composite fault detection and classification in wireless biosensor networks

Reliability is a critical aspect of wireless biosensor networks. In this context, an efficient methodology for fault diagnosis in wireless biosensor networks under composite fault scenarios is proposed. The methodology consists of three steps: firstly, hard fault detection in sensitive and non-sensitive regions using timeout response and Fletcher’s checksum implementation; secondly, soft fault detection through fault status generation using the Z-score test; and lastly, fault classification using a probabilistic neural network to categorize composite faults based on their behaviors. The proposed methodology is particularly well-suited for critical events in wireless biosensor networks. Hard fault detection is implemented in a biosensor network simulation setup, and its performance is evaluated in terms of packet delivery ratio and energy consumption, both before and after fault detection. For the hard fault detection, the proposed methodology improves the packet delivery ratio by $\sim$13.04% while reducing energy consumption by $\sim$11.96% in the sensitive region. In the non-sensitive region, the average biosensor node and link failure detection rate is $\sim$87%. Soft fault detection and classification are evaluated through simulations using human-body biosensor data and relevant fault evaluation metrics. Compared to its existing counterparts, the proposed methodology improves the detection rate by $\sim$8.81%, reduces the false positive rate by $\sim$33.25%, and reduces the false negative rate by $\sim$43.25%. For fault classification, the detection rate for permanent faults is $\sim$4.68% higher, and the misclassification rate is $\sim$45.09% lower as compared to other fault types. In addition, a T-score is performed to validate the statistical significance of the soft fault detection and classification results at a 95% confidence level. Experimental results demonstrate that the proposed methodology effectively detects and classifies composite fault scenarios, achieving superior performance compared to existing fault diagnosis methods.