A Method For Generating Process Topology-Based Causal Models
Process disturbances always spread along the connected equipment in a plant and are detected in many places. In order to identify the root disturbance, many data-based fault detection and diagnosis (FDD) methods have been developed in recent years. However, most of these methods can generate spurious solutions. Several authors have observed that FDD methods are enhanced if topology information about the causal relationships of a process is considered as well. Generally, this topology information is manually created by using the process knowledge. However, such a way is always time-consuming and the result is imprecise. Hence, there is a requirement for an automated generation of effective topology-based causal models.
This thesis developed a thorough approach to implement two types of causal models, i.e., a connectivity matrix and a causal digraph, based on piping and instrumentation diagrams (P&IDs). As the core development tools, AutoCAD P&ID and object-oriented programming (OOP) of MATLAB were used. The development included three procedures: generate topology data, define the class for generating causal models, and obtain the causal models by instantiating the class with the topology data.
In conclusion, it appears that both the connectivity matrix and causal digraph manifest the internal relationship between different process components caused by material flows and signal flows in a clear way. Therefore, these models can play an important role in the research associated with the FDD methods.
Thesis electronical version can be downloaded from here
This info last modified 20 Jun 2018 by Jukka Kortela