The high pressure LDPE (low density polyethylene) industrial process operates under supercritical conditions, and so it is necessary to monitor its performance to prevent abnormal situations. Extreme deviations from the normal operating region lead to conditions such as: loss of normal reaction, decompositions of the reactants, and lost production due to outages. Multivariate Statistical Process Control strategies operate on top of the DCS (distributed control system) to detect and diagnose abnormal process behavior and provide the operators an opportunity to take preventative operational actions. Process engineers may also use it for off-line diagnosis of poorly understood processes. In this work, data from a commercial LDPE/EVA (ethylene-vinyl acetate copolymer) high-pressure unit using an OPC (Object linking and embedding for Process Control) server installed on the DCS is used to build empirical models and perform fault detection. Data transfer issues, preprocessing, process model development using Principal Components Analysis (PCA) and first principles modelling of critical equipment are provided. In addition to showing grade transitions in the latent variable space, the models were used to detect process shifts. Process data from various real faults were considered and it was established that PCA could be employed to predict and diagnose process faults. The study gives recommendations for process monitoring strategies.