.png){kind=icon}
.png){kind=icon}
Return
Signal synthesis method for turbine blade crack detection based on vibration-acoustic simulation and actual signals confrontation
.png){kind=icon}
PRE
AI
X
X
R
D
F
DOI：10.1177/14759217261431149.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
<jats:p>With the advantages of noncontact detection, acoustic sensor has been widely applied to turbine for structural health monitoring and blade crack detection under harsh environment. However, it is difficult and time-consuming to obtain high-quality and sufficient monitoring sound signals, which lead to poor performance in crack detection based on limited sensor data. To obtain a large number of high-quality signals of turbine blade crack detection, the sound signal synthesis method is proposed based on vibration-acoustic coupling simulation (VACS) and actual signals confrontation. First, the VACS model is built for turbine from two aspects of vibration and acoustics, which can obtain high-quality and sufficient simulation sound signals. Besides, the conditional generative adversarial network (CGAN)-based signal quality improvement method is proposed to obtain generated signals based on actual signals confrontation with limited sensor data, which can bridge the gap between simulation and measured signals. Further, the CGAN model integration method is presented to synthetize sound signal by transferring generated signals from measured points to all simulation points, which are employed for turbine blade crack detection. To verify the proposed method, two case studies are implemented based on two types of turbines. It can obtain synthetic sound signals with over 0.68 cosine similarity, and the accuracy of blade crack detection can reach about 93% under different conditions. Comparing with other methods, the superiority and reliability of the proposed method is validated, which is significant for turbine acoustic data acquisition and blade crack detection.</jats:p>
Keywords:
vibration-acoustic coupling simulation
acoustic sensor
blade crack detection
conditional generative adversarial network
signal synthesis
Journal
S
IF:
0
Papers:
341
Citations:
0
Organization
Cited Papers