Computer-aided fault diagnosis based on the dynamometer card (DC) of the sucker-rod pumping system (SRPS) is a crucial technology to reduce operating costs and increase yield. Currently, the conventional method to implement this technology is the pattern recognition of the DC features. However, the training set of DC that determines the diagnostic accuracy of the method is difficult to obtain because of the differences between oil wells. Moreover, this method can only obtain the type of single fault without quantitative analysis, which may affect the formulation of the adjustment measures.

Therefore, in the present study, a quantitative diagnosis method independent of the training data is proposed. In order to obtain the operation process of SRPS under the comprehensive conditions of the fault, a simulation model involving fault effects is established according to the fault mechanism. Subsequently, the framework of the optimization inversion method is established to determine the fault parameters by minimizing the difference between the measured DC and the DC generated by the fault-mechanism model. Then, the strategy of the partition parallel optimization is utilized to improve the stability and efficiency of the inversion algorithm. Meanwhile, fifteen indicating parameters that directly reflect the type and degree of faults are defined.

Finally, the proposed diagnosis method is verified experimentally through the data of 20 actual wells. The obtained results demonstrate the effectiveness of the proposed method for diagnosing the types of single and coupling faults, as well as predicting the production rate.

基于抽油杆抽油系统（SRPS）测功机卡（DC）的计算机辅助故障诊断是降低运行成本和提高产量的关键技术。目前，实现该技术的常规方法是DC特征的模式识别。但是，由于油井之间的差异，决定方法诊断准确性的DC训练集很难获得。而且，这种方法只能得到单一故障的类型，无法进行定量分析，这可能会影响调整措施的制定。

因此，在本研究中，提出了一种独立于训练数据的定量诊断方法。为了得到故障综合条件下SRPS的运行过程，根据故障机理建立了涉及故障影响的仿真模型。随后，建立了优化反演方法的框架，通过最小化实测直流与故障机制模型产生的直流之间的差异来确定故障参数。然后，利用分区并行优化策略来提高反演算法的稳定性和效率。同时定义了15个直接反映故障类型和程度的指示参数。

最后，通过20口实井数据对所提出的诊断方法进行了实验验证。获得的结果证明了所提出的方法在诊断单一和耦合故障类型以及预测生产率方面的有效性。