Wireless Sensor Networks (WSNs) are networks formed using a large number of low-cost sensor nodes that have limited energy sources, limited processing capability, low storage capacity, and generate a large amount of sensed data with high temporal coherency. Due to high node density in sensor networks, the same data is sensed by many nodes, which results in data redundancy. The problem becomes worse if the redundant transmission contains both normal and faulty data. This creates the issue of differentiating between normal and faulty behavior. This redundancy can be eliminated by using data fusion based techniques. Data aggregation based data fusion is considered an important technique that can reduce the repetitive transmission of the sensed data and can improve the network lifetime. Hence for maintaining the reliability and longevity of the sensor network, in this article, we propose a novel combination of data aggregation based data fusion with effective fault detection by utilizing the properties of Grey Model (GM) and Kernel-based Extreme Learning Machine (KELM). Here, GM is utilized as a data fusion scheme that records the single datum pattern by rejecting the repetitive data received from the different sensor nodes. Trained KELM is utilized for effective detection of fault thus maintaining high confidentiality of the network. The proposed technique is trained and tested using the standard WSN datasets recorded from different laboratories. The simulation results show that the proposed technique can effectively reduce the repetitive transmission and can efficiently detect the fault in the network. The solved problems result in extending the lifetime of the network by taking the low computational time and fast speed.

无线传感器网络（Wireless Sensor Networks，WSN）是由大量低成本传感器节点组成的网络，这些节点具有有限的能源、有限的处理能力、低存储容量，并产生大量具有高时间一致性的传感数据。由于传感器网络中节点密度高，同一数据被多个节点感知，导致数据冗余。如果冗余传输包含正常数据和错误数据，问题就会变得更糟。这就产生了区分正常行为和错误行为的问题。这种冗余可以通过使用基于数据融合的技术来消除。基于数据聚合的数据融合被认为是一种重要的技术，可以减少感知数据的重复传输，提高网络寿命。因此，为了保持传感器网络的可靠性和寿命，在本文中，我们利用灰色模型 (GM) 和基于内核的极限学习机 (KELM) 的特性，提出了一种基于数据聚合的数据融合与有效故障检测的新组合。 ）。在这里，GM 被用作数据融合方案，通过拒绝从不同传感器节点接收到的重复数据来记录单个数据模式。训练有素的 KELM 用于有效检测故障，从而保持网络的高度机密性。所提出的技术使用从不同实验室记录的标准 WSN 数据集进行训练和测试。仿真结果表明，所提出的技术能够有效减少重复传输，能够有效地检测网络中的故障。