In the Wireless Sensor Network (WSN), the operation reliability is usually evaluated by processing the measured data at network nodes. As there are the problems of the complex calculation and large energy consumption in the traditional algorithms, a method for fault diagnosis of nodes in WSN based on particle swarm optimization and Gaussian distribution is proposed in this paper. Because of the characteristics of simple rules and fast convergence rate of PSO, the tested data are optimized and a corresponding threshold level is obtained. Then determine whether the node is good through judging whether the tested data meet the relationship with the threshold value range based on Gaussian distribution. The experimental results show that this fault diagnosis method can find the fault nodes promptly and effectively and improve the reliability of WSN greatly.