METHOD FOR THE IMPROVED DETECTION OF PROCESS ANOMALIES OF A TECHNICAL INSTALLATION AND CORRESPONDING DIAGNOSTIC SYSTEM

摘要

The invention relates in essence to a method and a diagnostic system for improved detection of a process anomaly of a technical installation, in which first a self-organizing map is trained using the historical process data as good conditions of the installation, wherein the condition for determining the time sequence or a Paths of the hit nodes and the tolerances of the hits of the neurons are used and wherein thresholds for the Euclidean distance for the Gutzustände be determined and stored, and in which the current process data of the system in the form of a state vector using the trained self-organizing map are evaluated, the Euclidean distance of the current state vector to the neuron hit is checked for an exceeding of the threshold value, wherein a neuron which should be taken is determined with the aid of the path determined, provided that the threshold value does not already exist Examining with the hit neuron has been exceeded and a symptom vector is determined from the current state vector and either the hit neuron or the neuron that should be hit. If optionally the number of hits of a node is stored together with the paths, then temporal deviations from the good state can be determined. The advantages are above all in an early and reliable detection of deviations from the normal state, the additional possibility and in the correct determination of a symptom vector to better find the cause of the error.