Wireless sensor networks(WSNs)are indispensable in modern social applications by virtue of their own advantages,and they are widely used in various fields.WSNs mainly consist of a large number of sensor nodes which are randomly deployed in the target area.Due to the small-size and fast-consumption sensors,they are prone to failure than other components.If most of sensor nodes suffer from fault,the network will disconnect and the network coverage will be incomplete.Moreover,the sensor nodes that produce some wrong data information may enable the manager make some wrong judgments.Therefore,it is vital to identify fault nodes in WSNs.In this paper,we propose some new fault diagnosis algorithms for different network conditions.The main contributions of this paper are as follows:(1)For the large-scale wireless sensor networks with stable changes of node data,we propose a fault-tolerant sensor diagnosis algorithm based on similarity.Firstly,according to the similarity of the historical data owned by each sensor node,the fault-tolerant diagnosis algorithm is established to evaluate the state of sensor nodes.When a node is considered as suspicious fault node,the algorithm sends a fault diagnosis request to neighbor nodes of the suspicious node.Then a fault diagnosis algorithm based on similarity is proposed to detect the final state of the sensor nodes.Algorithm analysis and experimental results show that,the algorithm proposed not only have a good performance in fault detection accuracy where the number of neighbor nodes is quite a few and the fault probability is high,but also greatly reduce energy consumption.(2)For the wireless sensor networks with any type of nodes data,we propose a fault diagnosis algorithm based on similarity and mobile sensor detector(MSD).By taking advantage of the Hilbert curve,our proposed algorithm designs an effective moving path for the MSD which performs the fault diagnosis for all sensor nodes by only one hop.For large-scale wireless sensor networks,we develop a fault diagnosis algorithm by multiple MSDs,moreover,we put forward a special feedback method to the sink node.The theoretical analysis shows that the proposed scheme can reduce the energy consumption and improve the lifetime of the network.The simulation results also show that our scheme is much better than existing two algorithms in terms of detection accuracy,false alarm rate,algorithm message complexity and energy consumption.(3)For wireless sensor networks where the environment is not suitable for mobile nodes,we establish a diagnosis algorithm based on similarity and fault-free node.The proposed algorithm adds some fault-free nodes in the network,and utilizes them to diagnose other nodes.Firstly,we utilize the k-means algorithm and the weight function to determine the location of the fault-free node,and then we propose the fault diagnosis algorithm using the fault-free node.Our algorithm can effectively reduce node energy consumption and achieve node energy balance.At the same time,it has a high fault detection accuracy and low false alarm rate for the network with high node failure rate,which meets the requirement of actual application for the wireless sensor network.