This MatLab-based software prototype expands the variable selection used in the FDI software. If only statistical relationships between measurement channels have been taken into account in the FDI software, the PIT software goes one step further. The software makes it possible to actually provide causal relationships between process variables.
Industrial processes and their interrelations within a system are usually very complex interrelated and therefore difficult to calculate mathematically. Also for long-time process experts very elaborate to recognize all mutual influences clearly.
On the one hand, FDI can be used to create more reliable models for fault diagnosis, thanks to improved variability selection. On the other hand, to provide an overview of the causal relationships in a process.
By means of the results of the PIT software, it is easier to recognize relationships and to increase the efficiency and reliability of a system with this knowledge. In order to identify the causal relationships, the PIT software essentially performs three steps:
- Formation of a basis of the correlated measurement channels by means of linear and non-linear correlation measures after data pre-processing.
- Application of variable selection algorithms on this basis.
- Execution of scheck correlation and causality analyzes, whereby the actually correlated measurement channels remain together with a cause-effect analysis in the data base.