A method of monitoring state of a system, comprising: providing a set of reference data (H) comprising a plurality of observations learned from sensors of a modeling system that characterize the dynamic behavior of the modeled system, wherein the reference data set (H) is in the form of a matrix, with each column of the matrix representing a point and each row representing one sensor values providing a current observation with regard to modeling system comparing the current observation with the reference data set (H) using a similarity operator for a similarity score for each observation learned in the reference data set (H) and if the similarity score for an observation learned is above a threshold or is one of a predetermined scores highest similarity in all learned observation number, including those observations learned in a subset of data (D) of the assembly reference data (H) calculating a model based on the current observation system and the current subset of data (D) derived from the set of reference data (H), wherein the calculation model comprises generating an estimation model comprising a weighted composite of the subset data (D), providing a series of subsequent current observations regarding the modeled system recalculating the model by further determining the subset of data (D) for each new current observation and detecting the incipiency of system failure testing model estimation in contrast to the current observation.Un procedimiento de monitorización de estado de un sistema, que comprende: proporcionar un conjunto de datos de referencia (H) que comprende una pluralidad de observaciones aprendidas a partir de sensores de un sistema modelado que caracterizan el comportamiento dinámico del sistema modelado, en el que el conjunto de datos de referencia (H) es en forma de una matriz, con cada columna de la matriz que representa una observación y cada fila que representa valores de un solo sensor proporcionar una observación actual con respecto al sistem
