Skip to Main Content
ASME Press Select Proceedings

Intelligent Engineering Systems through Artificial Neural Networks Volume 18

Editor
Cihan H. Dagli
Cihan H. Dagli
Search for other works by this author on:
ISBN-10:
0791802823
ISBN:
9780791802823
No. of Pages:
700
Publisher:
ASME Press
Publication date:
2008

Today's high cost of petroleum imposes much stricter requirements on aircraft engines for fuel-efficiency. Engines must operate within appropriate temperature ranges thus necessitating cooling of engine parts achieved through millions of laser-drilled holes in turbine engine blades and vanes. In order to maximize the benefits available from expensive laser drilling equipment, it is necessary to have the capability to predict from drill settings hole geometry and number of laser pulses required for puncturing material (“breakthrough”). There are no accurate and reliable analytic models available today for laser-drilled hole characterization, hence many laser drilling systems operate essentially on a trial and error basis. The work reported here discusses a method for prediction of acoustic emissions during laser drilling. Knowledge of acoustic emissions would permit control of both laser drill and hole geometry. Airborne acoustic emissions at the ablative surface monitored during laser drilling are used to construct a database which is used to train a backpropagation neural network to predict power spectral density of the acoustic wave given laser parameters such as pulse width, frequency and average power. Experimental investigations - performed using a P50 Nd:YAG Laser - involved drilling of holes in a Waspalloy steel plate and the calculation of power spectral density of microphone voltage for each laser pulse applied. Emphasis was placed on the acoustic emission associated with the first and second pulse applied at each hole since these produced the strongest signatures relative to background noise. Experimental results show a clear and consistent signature in the 0–10KHz range and confirm that the neural network can predict acoustic signatures with 96% accuracy. Additionally, the neural network gave valuable sensitivity information regarding which laser parameters were the most significant for acoustic emission PSD.

This content is only available via PDF.
Close Modal
This Feature Is Available To Subscribers Only

Sign In or Create an Account

Close Modal
Close Modal