Dynamic risk assessment and fault detection using a multivariate technique


  • This work was supported by Vale Inc., Research & Development Corporation (RDC), and the Atlantic Canada Opportunities Agency (ACOA) under the Vale Research Chair Program.


In the context of process safety, significant improvements are needed in fault detection methods, especially, in the areas of early detection and warning. In this article, a multivariate risk-based fault detection and diagnosis technique is proposed. The key elements of this technique are to eliminate faults that are not serious and to provide a dynamic process risk indication at each sampling instant. A multivariable residual generation process based on the Kalman filter has been combined with a risk assessment procedure. The use of the Kalman filter makes the method more robust to false alarms, which is an important aspect of any fault detection algorithm that targets the safety of a process. In addition, we consider significant differences in the severity of the faults associated with different process variables. We also take into account the varying intensity of damage caused by the increasing and decreasing rates of fault and the need to treat those cases differently. © 2013 American Institute of Chemical Engineers Process Saf Prog 32: 365–375, 2013