Improved correlation analysis and visualization of industrial alarm data

► A Gaussian kernel method is applied to convert binary alarm data into pseudo continuous time series. ► We use a correlation color map to visualize the time-lagged correlation between different alarms. ► Alarm tags are reordered to better group the correlated alarms to find redundancy in alarm settings. ► Multivariate analysis based on the pseudo data can reduce the influence of missed and false alarms.