Method and device for automatically determining alarm threshold values

Abstract

The subject of the application is a method and device for automatically determining alarm threshold values. It applies specifically to machine and process monitoring systems used in industry, e-health, defense, and other fields where numerous data are collected from various sensors. The method involves a data acquisition module (2) collecting time series from sensors (1) and the parameters of these series online. The determined time series are saved to a historical database (3). In offline mode, after installing a new system, and then upon user request, the Data Analysis Module (4) is launched. Its task is to automatically determine alarm threshold values. If it is possible to define the machine's operating states, thresholds are determined separately for each state.The Data Analysis Module consists of the following elements: a) removing outliers from the time series (5), b) modeling the time series (6) using at least one statistical distribution, c) determining the baseline value (7) by specifying a selected level of the distribution function, d) determining the threshold value (8) for the warning and alarm states by applying a logarithmic transformation to the modeling results. The determined values ​​are saved in (9). In online mode, the system periodically measures the input values ​​- module (2), and then activates the Alarm Detection Module (10), which checks whether the measured values ​​exceed the threshold values ​​(9). If the beginning or end of an exceedance is detected, the system sends information to the Interface Module (11), which enables transmission of information to other systems, visualization of results, and manual adjustment of alarm threshold settings.