This research focuses on the comparison of the fault detection and diagnosis performances achieved by two different modeling techniques, PLS and N4SID, applied to the data generated by the TE Simulator. For each model, a number of faulty datasets are monitored using the T2 and the Q statistics and, once an out-of-control signal is obtained, the fault diagnosis is performed using contribution plots and cumulative sum plots for regression adjusted variables. The results of each technique are confronted with expected theoretical results, and performance indexes are generated. These overall results for the PLS and the N4SID models are then compared, leading to the conclusion that the performance of the N4SID model is, in terms of detection time and diagnose power, superior to that of the PLS model. In addition, a maintenance strategy is derived based on the fault monitoring and diagnosis results, focusing on the production cost of the plant.