Data set processing for the optimization of the artificial intelligence-based diagnostic methods |
|---|
| Piotr Bilski |
- Abstract:
- The paper presents the application of selected statistical methods to process the training and testing data sets and prepare them for the artificial intelligence-based method used in the diagnostics of analog systems. The size of the set (especially the number of processed attributes – stamps) determines the efficiency of the selected algorithm and minimizes the amount of information to be measured in the actual system. The preprocessing operations include elimination of constant or quasi-stationary stamps and selection of their most important set, allowing for the efficient fault detection or parameter identification. The paper presents two methods from the econometrics domain adjusted to the technical diagnostics applications. Their implementation is tested on the electronic analog filter. Also, efficiency of the artificial neural network (ANN) working with the original and preprocessed data is verified.
- Download:
- IMEKO-TC10-2013-026.pdf
- DOI:
- -
- Event details
- IMEKO TC:
- TC10
- Event name:
- TC10 Workshop 2013
- Title:
12th IMEKO TC10 Workshop "New Perspectives in Measurements, Tools and Techniques for Industrial Applications"
- Place:
- Florence, ITALY
- Time:
- 06 June 2013 - 07 June 2013