SYSTEM AND METHOD FOR DETECTING AND MEASURING ANOMALIES IN SIGNALLING ORIGINATING FROM COMPONENTS USED IN INDUSTRIAL PROCESSES Russian patent published in 2022 - IPC G05B23/02 

FIELD: measuring.

SUBSTANCE: invention relates to a method and system for detecting anomalies in sensor data originating from components used in industrial processes. The method includes the following stages: - measuring and/or tracking the measurement data, accordingly, tracking the process parameters (4) of components used in the industrial process (6) by means of measuring apparatus or sensors (2), and identifying same-sized time frames in the measurement and/or process parameters (4) for the time frames wherein the components used in the industrial process (6) function normally, wherein the measurement and/or process parameters (4) contain parameter values for multiple measurement/sensor process parameters (41) and/or variables (42); - converting the values (4) of the parameters of the multiple measurement/sensor process parameters (41) and/or variables (42) into observable binary processing codes for each of the identified same-sized time frames, and assigning binary processing codes to the sequence of stored Markov chain states; - forming a multidimensional data structure containing a set number of values of the variable parameters of the hidden Markov model, wherein the variable parameters of the multidimensional data structure model are determined by means of a machine learning module (8) applied to the sequence of stored Markov chain states with the assigned binary processing codes (91), and wherein the variable parameters (811, 812, …, 81x) of the hidden Markov model of a multidimensional data structure vary and are trained by training the normal state frequency (82) of the occurring emergency events based on the measurement data and/or process parameters (4) of the identified same-sized time frames; - initialising and storing multiple values (83) of the probabilistic state by applying the trained multidimensional data structure with the values of the variable parameters of the hidden Markov model to the pre-discretised binary processing codes with same-sized time frames identical to the time frame for the values (4) of the parameters of the multiple measurement/sensor process parameters (41) and/or variables (42); - determining the logarithmic threshold value of the quantitative indicator of the anomalies by coordinating the logarithmic resulting values of the stored values (83) of the probabilistic state; and - deploying said trained multidimensional data structure with the values of the variable parameters of the hidden Markov model in order to track new measured, accordingly, determined measurement data and/or process parameters (4) from industrial equipment or factories (1) using the threshold value of the quantitative indicator of anomalies so as to detect abnormal values of the sensor data potentially indicating the future system failure, wherein, when the values of sensor data are abnormal, the logarithmic resulting value for the value (83) of the probabilistic state of the new measured, accordingly, determined measurement data and/or process parameters (4) for initiation is formed and compared with the stored values (83) of the probabilistic state based on said logarithmic threshold value of the quantitative indicator of anomalies.

EFFECT: equipment of the industrial enterprise system with a detection and signalling apparatus based on machine learning tools triggering sensors and measuring equipment of the industrial enterprise.

11 cl, 16 dwg