OPTIMIZING OF PROCESSES BASED ON SELF-ORGANIZING MAPS

摘要

It proposes an optimization method for the data-based optimization of continuous or discontinuous processes that take place on an industrial plant, in particular manufacturing or production plant, where historical data can be retrieved and evaluated from processes which have already expired, comprising: acquisition of measured variables, decision variables and measurable disturbances within the framework of the processes can be tapped; Teaching the previously acquired data into a self-organizing map, wherein nodes of the map represent individual passes of the processes; Specification of a quality criterion; Calculation of the quality criterion with respect to each node of the self-organizing map; Perform a new run of the process and capture the current measurable disturbances and measures; Identification of the node within the self-organizing map whose stored disturbance variables and measured variables have a minimal difference to the newly detected disturbance variables and measured variables, and whose quality criterion has the smallest possible value, or whose quality criterion has a minimum value, and its stored disturbance variables and measured variables have the smallest possible difference to the newly detected disturbance variables and measured variables.