Neural network technology for processing of signals from technical objects control means

The paper presents a technology based on an ensemble of neural networks that solves the problem of predicting a multidimensional time series of telemetry data. A distinctive feature of this is the possibility of additional training in the case of analyzing parameters in a non-stationary environment, providing operational processing of the entire set of telemetry parameters of the technical object, increasing the accuracy of identification and prediction of the state of the object by telemetric information with incomplete and noisy input data. This reduces the costs of monitoring the status and behavior of the control object.