Dynamometer cards are commonly used to analyze down-hole working conditions of pumping systems in actual oil production. Nowadays, the traditional supervised learning methods heavily rely on the classification accuracy of the training samples. In order to reduce the errors of manual classification, an automatic clustering algorithm is proposed and applied to diagnose down-hole conditions of pumping systems. The spectral clustering (SC) is a new clustering algorithm, which is suitable for any data distribution. However, it is sensitive to initial cluster centers and scale parameters, and needs to predefine the cluster number. In order to overcome these shortcomings, we propose an automatic clustering algorithm, fast black hole–spectral clustering (FBH–SC). The FBH algorithm is used to replace the K-mean method in SC, and a CritC index function is used as the target function to automatically choose the best scale parameter and clustering number in the clustering process. Different simulation experiments were designed to define the relationship among scale parameter, clustering number, CritC index value, and clustering accuracy. Finally, an example is given to validate the effectiveness of the proposed algorithm.