A data analysis and signal based system identification tool
GPMAS “General Purpose Mathematical Application Server” is a MATLAB based tool for data analysis and model identification.
The user can perform data pre- and post-processing and use several black box model techniques for identifying the best-suited model for the specific application.
In addition the identified models parameter can be automatically exported as they are compatible with the corresponding models, compiled as DLLs.
In particular the GPMAS allows:
- pre-processing signal analysis which include frequency analysis, filtering, outlier removal, etc.
- models identification, based on black-box tech-niques which include neural networks, state-space, observer based models, etc.
- validation and testing of customized fault detec-tion algorithms.
Within the GPMAS environment is possible to:
- load measured data coming from one or several tests on the same or different plants,
- perform a signal pre-processing (data synchro-nization, smoothing, de-trending, filtering and units conversion),
- combine signals together by custom made for-mulae for creating new ones to be used during identification (e.g. power using revolutions per minute and torque),
- perform system identification using several techniques, based on:
- Signal Based techniques
- Causal Graph approach
- Observer based models
- State Space models
- Multi-Layer Perceptron Neural Networks models
- Radial Basis Functions Neural Networks models
- Dynamic Functional Link Neural Net-works models
Input data formats are specified by an XML (eXten-sible Markup Language) file, and provided either ASCII or MAT format or for the data signals.
In the GPMAS environment the user can save at any time the current session, that can be later re-loaded to proceed with the work exactly from what it was stopped.
Besides, the processed signals can be saved in MATLAB figures files (.fig).
As part of this tool SATE developed for the implementation of the identified models, compiled code in the form of DLLs for the most common models.
These DLLs are compiled diagnostic libraries implementing generic functions or models which can be called by any software program with appropriate documented interface (i.e. input/output) procedures and syntax.
According to the system/process to be modelled, the models parameters will change but the code implementing such models will be the same.
These DLLs can be used by any device supporting DLLs technology, e.g. Windows based platforms.
The parameters exported from the GPMAS identification session are already fully compatible with these DLLs.