Sensor nodes in a wireless sensor network are influenced by the surrounding environment while monitoring data, which can lead to faults and data biases, resulting in erroneous decisions and losses. Identifying and classifying fault types is a challenge that still need to be addressed. The inertia weight \(\omega\) and learning factor c were optimized to enhance the optimization ability of the particle swarm. Additionally, parameters such as \(\sigma ,\textrm{n} ,d,\gamma\) in the hybrid kernel function, and the penalty coefficient C, were also optimized to improve classification accuracy. A diagnosis model of a hybrid extreme learning machine based on an updated particle swarm optimization for sensor node faults was developed. The dataset of water parameters was obtained based on constructing a monitoring system for intensive aquaculture. Four different proportions of fault data, respectively 5 \(\%\) ,10 \(\%\) ,15 \(\%\) , and 20 \(\%\) , were added to the dataset to create new datasets for training the model.Test results of the new diagnostic model show an average classification accuracy of 99.30 \(\%\) ,indicating that the proposed fault diagnosis model in this study enhances fault classification accuracy compared to other diagnostic algorithms.
