Location
FIU
Document Type
Event
Keywords
Analog Neural Network, Signal Processing, Surface Response Excitation, Fast Fourier Transform, Polylactic Acid, Digital Oscilloscope, Function Generator, Levenberg Marquette Algorithm, Structural Health Monitoring, Deep Learning, and Machine Learning
Description
It is possible to create voids inside polymer parts built by additive manufacturing methods. In this study, the feasibility of using structural health monitoring (SHM) methods was evaluated for identification of the voids in polymer parts. The Surface Response to Excitation (SuRE) method was used to collect the experimental data.
Ten Polylactic Acid (PLA) bars of identical external geometries were manufactured using Fused deposition modeling (FDM) type 3D printing. Four bars were solid without any voids inside of them. Two bar groups were manufactured with the same void lengths of 1 mm, 2 mm, and 3 mm at the center. Piezoelectric elements were attached to the opposite ends of all bars. One piezoelectric element was excited with 20 pulses. Each pulse had different widths. The monitored signal at the other piezoelectric element was acquired by using a digital oscilloscope. Fast Fourier Transformation (FFT) of the signals was calculated. A Levenberg Marquette type neural network was trained with the data of the four bars. These bars were the one without any void and three others each with one of three void lengths. The inputs of the neural network were 10 averages of different frequency ranges at the calculated spectrums with the FFT. The neural network had one output. The output was 0 for no void and 1 if the bar had a void. Trained neural network inspected six other cases which were not used during the training. The neural network correctly classified more than 90% of the cases.
Study indicated that SuRE method with pulse excitation and neural network-based classification is capable of identifying internal voids.
DOI
https://doi.org/10.5038/MSUO7502
Detection of Internal Voids of Additively Manufactured Parts with Structual Health Monitoring Methods
FIU
It is possible to create voids inside polymer parts built by additive manufacturing methods. In this study, the feasibility of using structural health monitoring (SHM) methods was evaluated for identification of the voids in polymer parts. The Surface Response to Excitation (SuRE) method was used to collect the experimental data.
Ten Polylactic Acid (PLA) bars of identical external geometries were manufactured using Fused deposition modeling (FDM) type 3D printing. Four bars were solid without any voids inside of them. Two bar groups were manufactured with the same void lengths of 1 mm, 2 mm, and 3 mm at the center. Piezoelectric elements were attached to the opposite ends of all bars. One piezoelectric element was excited with 20 pulses. Each pulse had different widths. The monitored signal at the other piezoelectric element was acquired by using a digital oscilloscope. Fast Fourier Transformation (FFT) of the signals was calculated. A Levenberg Marquette type neural network was trained with the data of the four bars. These bars were the one without any void and three others each with one of three void lengths. The inputs of the neural network were 10 averages of different frequency ranges at the calculated spectrums with the FFT. The neural network had one output. The output was 0 for no void and 1 if the bar had a void. Trained neural network inspected six other cases which were not used during the training. The neural network correctly classified more than 90% of the cases.
Study indicated that SuRE method with pulse excitation and neural network-based classification is capable of identifying internal voids.