Objective 1 To show that through the application of analysis with AI techniques and statistical methods, it is possible to successfully address highly complex problems of multiple variables and nonlinear functions involved in refining and petrochemical processes, facilitating the identification of one or more problem variables as well as hidden patterns that affect these processes, In this research we were able to determine the temperature of sublimation of salts at high pressure of a plate heat exchanger and the realization of a novel procedure for the industry, for cleaning the exchanger in operation, without the need for plant shutdowns, reducing operational and maintenance costs. Objective 2 To prove that the development of Artificial Intelligence based models applied to refining processes is possible to analyze large volumes of sequential data to enable fast and assertive response and facilitate timely, safe and reliable decision making. Objective 3. To demonstrate how AI in the processes of the Ecopetrol refinery (success case) identified the problem variable of a complex physicochemical phenomenon without any apparent cause, such as the increase in the differential of a plate heat exchanger, specifying that the rate of increase depended mainly on the inlet temperature of the hot side of the product, which impacted the optimization of the operation.