An Efficient Approach to Detect Faulty Readings Using Fuzzy Logic and Risk Number: Applicability in Long-thin Wireless Sensor Networks

Wireless sensor networks (WSNs) consist of hundreds or thousands of tiny sensor nodes that work together to do some special tasks. Detecting node with faulty readings is one of the important issues in WSNs. The weighted voting is usually used to detect faulty readings. Most of the existing algorithms use correlation of two nodes read vectors as weight to detect faulty readings, which result in high computational complexity. In addition, algorithms that use inverse of distance as weight have weaknesses in accuracy of calculations. This paper proposes a new fuzzy-based algorithm to detect faulty readings. Using an effective fuzzy inference system can improve the decision-making algorithm, which is used for detecting faulty readings in WSNs. We use normal and weighted averages of read vectors sensed values instead of single value or the entire of these vectors. The computation complexity of our algorithm is low and the accuracy is at a high level.