According to the project proposal, PIMMS is aiming at developing a diagnostic system, which will be applicable to a diverse range of equipment components, as well as cutting processes. The inherent difficulties with this are:
All measurements collected so far gave strong evidence in support of the above observations. This makes it impossible to develop diagnostic models, which would exhibit consistently good behaviour in all circumstances.
An important question is whether instead it is possible to "transform" each individual set of monitored parameters, to a common numerical representation, which would allow a single adaptive computation system to capture the individual characteristics of each different problem. That could in principle be made possible if a robust fault tree / symptom tree architecture can be developed for different scenarios, followed by a pre-processing module which will bring the different sets of monitored parameters to the desired common numerical representation. Therefore, critical to the success of the PIMMS system will be the development of:
The information then can be passed to an adaptive computation processing, which would consist of
The overall architecture will rely on the development of appropriate user interfaces to support the individual processing modules, as well as suitable software interfaces between the individual modules.
The first Block Diagram of this document shows an example of how information can be exchanged between fault and symptom tree modules, a fuzzy information pre-processing module and the adaptive computation modules. The second diagram illustrates the interface between the adaptive computation module and the fuzzy information pre-processing module, as well as the data processing flow in the Novelty Detector / Classification modules.