Previous Page | Noesis Main Page | Next Page

Noesis Unsupervised Pattern Recognition (UPR)

 

 

Noesis offers very advanced functions in acoustic emission or arbitrary data analysis and clustering (data grouping). Unsupervised Pattern Recognition (UPR) is the process by which mathematical algorithms and neural networks are used to separate the data set (all data) into groups (clusters) which contain similar data. The data are grouped as similar depending on their features and a number of user choices. The user can select the features to be used for the data clustering the method to be used (algorithm) and several other parameters which can control / improve the method. The results can provide an insight to the physical phenomena producing each type of emission. The various clusters are shown in different (user defined) colors and labels and statistics are calculated for each cluster (class). All actions are easily undone to provide a high level of flexibility and user friendliness. The following is a list of the functions available with UPR in Noesis:

• UPR Wizard lets even inexperienced users perform complex UPR algorithms. The wizard provides information about pre-processing, UPR methods and method parameters and guides the user.

• Data pre-processing, feature selection, normalizing, projection generation etc. to assist in more efficient and arithmetically solid clustering via UPR.

• Automatic pre-processing of any data set.

• Multiple UPR algorithms, including Neural Networks, for automatically clustering data (Max-Min Distance, k-Means, LVQ Net etc.).

• All actions are applied to a Working Copy of the data leaving the Main Data Set unaffected for better result viewing and reporting.

• Manual clustering is available for evaluation and classification using common AE practices (see also Data Handling).

• Classification result output to PAC (DTA, TDA or WFS) files (see also Data Handling).

• Descriptive statistics regarding classification (see also Statistics).