Validity identification and classification technique of tank acoustic emission testing signals based on clustering analysis

As a modern no-monitoring identification technique, clustering can be used to classify data and reveal its internal structure under the no-experience knowledge condition. Applying floating threshold to re-calculate common feature parameters based on the acoustic emission (AE) waveforms data, the input vectors of clustering algorithm are obtained. With optimized K means clustering algorithm and obtained vectors, clustering effect is significantly improved. Through applying this method on tank AE inspection data, the result shows that different type acoustic sources and different propagation route sources can be distinguished with the achieved method. Also, good denoising effect is obtained. With these, tank floor AE testing and evaluation accuracy is improved.