The dynamometer cards (DC) are the data shown as closed curves collected from Sucker Rod Pumps, which are essential evidence to monitor the working states in modern oil are engineering. To meet the actual needs of oil fields, recently, the computer-aided diagnosis techniques are becoming useful measurements to help engineers monitoring the wells. Nevertheless, how to collect the various kinds of fault data from a well is always a puzzle for the application of the computer-aided methods, because of a well hardly experiences many types of faulty working states. The typical solution for this problem is building an album containing DCs collected from different wells, but this approach neglects the property differences between wells, which may influence the diagnosis accuracy. In order to address this tough issue, in this paper, a novel approach regarding generating DCs is proposed based on the analysis of the mechanism of a sucker rod pump (SRP) at normal and several faulty scenarios. This method could use the productive parameters and operation rules of a well to calculate the DCs at different working states based on dynamic mechanism analysis. Subsequently, according to the data support of generating DCs, the Hidden Markov Models under a specifically designed framework is used to build the relationships between DCs and working states. At last, the proposed method is verified experimentally through the productive parameters of many wells collected from an oilfield, and then some conventional techniques are employed in the comparison studies. The obtained results demonstrate the effectiveness of the proposed method for diagnosing the working states of Sucker Rod Pumps. (C) 2019 Elsevier Ltd. All rights reserved.