This dataset consists of AE signals generated by different sources on a PMMA plate. It is intended for studying and experimenting with data-based approaches for AE source classification.

The PMMA plate is excited using three distinct AE sources:

1. Pencil-lead breaks (Hsu-Nielsen source with 0.5 mm leads)
2. Pulsing by the AMSY-6 acoustic emission measurement system with 30 Vpp
3. Salt trickling from a salt shaker, positioned approximately 20 cm away

YouTube Video

Structure

• Material: Polymethyl methacrylate (PMMA)
• Dimensions: 300 x 300 x 9.7 mm

Measurement chain

• Vallen AMSY-6 AE measurement system with ASIP-2A signal processor cards
• 4x Vallen VS150-RSC sensors with 34 dB preamplifier gain
• Measurement settings:
  • Filter: 95-300 kHz
  • Threshold: 40 dBAE
  • Duration discrimination time: 1000 µs
  • TR (transient/raw data):
    • Samplerate: 2500 kHz
    • Pretrigger samples: 100
    • Post-duration samples: 1000

Sensor positions

The sensors are arranged in a square configuration on the PMMA plate:

Files

Each measurement consists of two SQLite3 database files:

• *.pridb: The primary database, that stores just the AE features (amplitude, energy, counts, duration, rise time, ...) and the index of the transient/raw data TRAI. Use the table view_ae_data.
• *.tradb: The transient database, that stores the raw data for each hit record referenced by the TRAI index. Use the table view_tr_data. The raw data is stored in the Data column in binary format (int16 values). The factor from column TR_mV can be used to transform the integer values to millivolts.

You may want to use the Python library vallenae to easily read the measurement files.

Examples

• classifier-openae.ipynb: SVM classifier trained with OpenAE features
• classifier-standard.ipynb: SVM classifier trained with standard AE features