New, faster, better transient recorders generate large amounts of waveform data. Looking at data, waveform by waveform, is not an efficient procedure. A tool is needed that recognizes similarities and differences in AE-signals and makes them available for statistical analysis: VisualClass™.

Look at these signals. You clearly see similarities as well as differences, but how can such information held inside the waveforms be extracted?

VisualClass™ recognizes the similarities and differences in AE-waveforms and develops a classifier that assigns each waveform to one of several classes, each with its own class number. Further information, such as how well a waveform fits into the assigned class, is also generated in terms of distance-ratios. All these classifier results can be linked to the corresponding hit-data in VisualAE™. This means, the classifier results can be statistically processed, graphically or numerically displayed or used for filtering like any traditional AE feature.

Highlights:

• Development environment to create classifiers for use with Vallen AMSY-5, AMSY4, and AMS3 transient data files
• Fully transparent by many diagrams showing current classifier training state
• Learning strategies optimized for Acoustic Emission signals  
• High performance un-supervised learning strategies for automatic cluster detection
• On-line context sensitive help for all menus
• "Getting started" tutorial including sample data
• Separate classifier program for quick use of classifiers exported by VisualClass™
• Classification results can be displayed along with conventional parameters in VisualAE™
• Additional results provide information about class fitting quality for each classified pattern
• Utility programs for most comfortable building of the prototype database:
• Waveform selection by a script-file, which can automatically be generated by the multi-plot program allows filtering according to class distances, hit data, parametric data, locations etc.
• Manual classification of prototype waveform data via drag-and-drop.