The process industries have increasingly focused on the domain of agile manufacturing to increase their competitiveness globally. Agile plants operate in a number of modes and frequently switch between them. The process of mode switching is termed as process transition. This paper seeks to optimize transition operations by developing a methodology to identify and classify different types of transitions from continuous production data. Traditional data analysis methods perform poorly on multi-state temporal signals, so, a new method based on principal component analysis is proposed for transition classification. By analyzing previous plant operating data, different instances of a transition can be identified. Good operating strategies are then extracted by comparing the instances. A self-organizing map based method is also proposed for visualization of process transitions. We illustrate the proposed transitions classification and performance analysis methods by application to a refinery hydro-cracker.