Change-rate-based Univariate Alarm System Design

With the rapid development of modern industry,the role of alarm systems in the monitoring of production processes has become increasingly significant.Alarm systems are used to detect abnormal conditions of industrial signals,communicate abnormal conditions to operators,and remind operators to deal with abnormal conditions in time.It plays critically important roles for the safe and efficient operation of industrial plants.Traditional univariate alarm systems rise alarms by comparing the amplitude of a process variable with an alarm threshold.In the current industry,there is a lot of signals that need to be judged whether the system is abnormal according to the rate of change(rather than amplitude changing),such as the change in equipment temperature and the main steam pressure of the steam turbine.Therefore,a new change-rate-based univariate alarm system is proposed in this paper,which can determine the abnormality of the system by calculating the change rate of the process signal.At present,there are still two unresolved problems in the practical application of the alarm system:1)How to select training data from huge amount of historical data to calculate the alarm system parameters?How to estimate the true value of the monitoring signal faster and more accurately when the system is affected by noise?2)In order to solve the above problems,the design method proposed in this article mainly includes the following innovative contents:(1)In this paper,historical data is selected by judging the probability distribution stability of the data.The probability distribution stability test of the historical data is based on the Bayesian estimation method.The parameters of the alarm system,namely the slope alarm threshold is obtained by selecting the stable and appropriate training data and the set of historical slope with stable probability distribution.(2)In this paper,the method of piece-wise linear representation is used to determine the segments of linear pieces of data in the sliding window,which is represented by linear equations.The influence of noise is reduced as much as possible,and the true value of the analog signal is estimated more quickly and accurately.It is more convenient for operators and analysts to extract valuable and hidden information from huge amount of historical data.The difficulty is to determine the number of segments for the piece-wise linear representation method.Therefore,a method is proposed to determine the number of segments based on root mean square error.The number of segments is determined with higher calculation efficiency,which ensures the real-time performance of the online operation of the alarm system.Numerical and industrial examples are provided to illustrate the research results of proposed method.The graphical user interface is designed to facilitate users and validate the results via visualization.The research results show that the application effect of the change-rate-based univariate alarm system is consistent with the design goals,and the effectiveness of the change-rate-based univariate alarm system and the method for determining the probability distribution stability is validated.