Determination jump monitored parameter using a neural network

Abstract

Model and algorithm of warning about the dangerous change in the parameter of the technical object designed. The algorithm is based on the diagrams constructed and operates in real time. Local array of time series points characterizing parameter chart forms. Each point on the graph the current value of the parameter and the following parameter value is formed. The time window is determined first. Array cut time window that moves along the time series. The sensor data in the process of forming a time series are used. Determination of dangerous changes in the parameters is carried out using a modified neural network.