Well failures in oil field assets lead to production loss and can greatly increase the operational expense. Predicting well failures before they occur can dramatically improve performance, such as by adjusting operating parameters to forestall failures or by scheduling maintenance to reduce unplanned repairs and to minimize downtime. This paper presents failure prediction framework and algorithms for rod pump artificial lift systems. It adapts state-of-the-art data mining approaches to learn patterns of dynamical pre-failure and normal well time series records, and to use these patterns to make failure predictions. Also, we develop a semi-supervised learning technique, called "random peek", in order to adjust the training process to cover more feature space and to overcome the bias caused by limited training samples. The data set from this paper is taken from a real-world asset using rod pump artificial lift systems. The results show that the failure prediction framework is capable of capturing future rod pump and tubing failures.