Root-cause Analysis Of Industrial Alarms Based On Multiple Linear Regression Models

Industrial alarm systems play important roles in ensuring safe production and efficient operation of modern industrial plants,and root-cause anaylsis of the alarms is one of the vital subjects in the field of industrial alarm systems.The root-cause analysis of the alarms can help operators to identify the root-causes of the abnormality and avoid major safety incidents.Therefore,it is very important to ensure safe operation of industrial systems to find root-causes of alarms correctly.From the trend characteristics of variables,this thesis proposes a method for analyzing the root-cause of alarms based on multiple linear regression models.Firstly,the piecewise linear representation method is used to extract the qualitative trends of monitored variables,and a method for determining the optimal number of segments for qualitative trends based on the goodness of fit is proposed in this thesis.The proposed method can ensure that each segment of monitored variables contains only one qualitative trend,which can help operators to determine the characteristic relationship of monitored variables.Then accurate qualitative trends of monitored variables will be provided for root-cause analysis of alarms.Secondly,the correlation between each variable and the principal variable is calculated by combining the piecewise linear representation method and Pearson correlation coefficient when the principal variables change significantly in the historical data.Then the variables that have a strong correlation with principal variable will be provided for root-cause analysis of alarms.Finally,multiple linear regression models of associated variables and principal variable are established to induce the contribution formula,and the root-cause can be determined based on the amount of contribution.Numerical examples are provided to illustrate the proposed method.And the proposed method is also applied to the root-cause analysis of industrial alarms such as the condenser of thermal power generation units.The root-cause analysis results show that the proposed method can correctly find the root-causes,which further prove the effectiveness of the proposed method.