Abstract: Aiming at the shortcomings of the conventional classification method of structural plane production， a new structural plane classification algorithm was proposed. Based on the structural plane classification of K-Means algorithm， the AFSA-RSK structural surface classification algorithm is established by combining the artificial fish swarm algorithm(AFSA)with the K-Means algorithm. The powerful optimization ability of the fish swarm algorithm is used to replace the K-Means algorithm to search for the structural surface set， and clustering by K-Means algorithm. After the clustering is completed， the corresponding parameters are selected to evaluate the clustering effect. According to the existing problems， the step size and visual field of the fish swarm algorithm are modified， the accuracy of finding the cluster is improved， and the clustering process is dynamically adjusted. Comparing the improved AFSA-RSK algorithm with other algorithms， it can be obtained that the improved AFSA-RSK algorithm is better in iterative speed， clustering precision and memory ratio， and it is more suitable for application in structural plane grouping.

Key words: artificial fish swarm algorithm(AFSA), rock joint plane, rock mass, clustering;slope