An autoclave reactor offers flexibility in producing specialty grades of ethylene/vinyl acetate copolymer at a temperature range of 150-230 degrees C and pressure range of 1400-2000kg/cm(2). At such conditions, copolymerization is accompanied by decomposition of reactants which increases the risk of thermal runaway. The runaway reaction is initiated by local hot spots in the reactor which are generated by process/equipment disturbances or imperfect mixing in the reactor. It is hard to predict decomposition, due to the extremely fast dynamics of the event. A decomposition detection method was developed based on the overall energy balance around the reactor. During normal steady state operations, the heat balance error should be within reasonable limits. If abnormal conditions generate excess heat in the reactor, the heat balance error will exceed the limit, indicating the possibility of an impending decomposition. Principal component analysis (PCA) was used for model identification and to get the unknown parameters in the model. An iterative PCA technique was used to confirm the selection of the model. Long term plant operation data over a period of six months were used for training and testing of the model. Data validation rules were applied and false alarms associated with operating condition fluctuations were minimized by applying appropriate equations for conversion at various operating conditions. The model was validated with an actual plant steady state decomposition case where the model could predict decomposition with a few seconds of lead time.