AE05.pdf

Self organizing map application may be found in a

This preview shows page 27 - 28 out of 39 pages.

self-organizing map application, may be found in a neural network tutorial. 51 Procedure Tensile Test The composite tensile test object used to gather the acoustic emission test failure mechanism data was a wet layup composed of eight layers of unidirectional fiberglass epoxy oriented along the tensile axis at 0 rad (0 deg) and manufactured in accordance with a published standard. 52 Aluminum tabs were bonded to the ends using the matrix epoxy to prevent grip noise. A transducer having a 150 kHz resonant frequency was oil coupled and attached to the center of the test object with electrical tape. An acoustic emission analyzer was used to record and store the parameter data from the test. The acoustic emission test equipment settings were as follows: 40 dB preamplifier gain with a 100 to 300 kHz bandpass filter, 40 dB threshold, 20 dB system gain, 40 μs peak definition time, 150 μs hit definition time and 300 μs hit lockout time. The tensile test machine was set for a grip pressure of 10 MPa (1400 lb f ·in. –2 ) and a load rate of 37 N·s –1 (500 lb f ·min –1 ). 53 Neural Network Application A kohonen self-organizing map neural network (Fig. 21) mapped the six- dimensional acoustic emission feature data — rise time, counts, energy, duration, amplitude and counts to peak — from each event into a single X, Y coordinate on a two-dimensional plot. Note here that, although the processing elements or neurons are interconnected between layers, there are no connections within layers. Self-organizing map training was accomplished using the acoustic emission feature data from every tenth acoustic emission hit throughout the loading ramp for input. Once trained, the network was used to classify the acoustic emission feature data from the remaining nine tenths of the data set. Results Classification of Failure Mechanisms Varying sizes of self-organizing map processing layers were implemented for classification: 20 × 20, 10 × 10, 5 × 5, 3 × 3 and 2 × 2. To begin with, it was found that the 20 × 20 layer provided far too many classification possibilities for the input data. This resulted in a two-dimensional plot of X, Y output with no distinctive features. The features only become apparent by generating a three-dimensional plot of frequency occurrence versus X, Y location (Fig. 22). Here, the input data are seen to be clustered into multiple peaks in what appear to be a series of five mountain ranges. Similar results were obtained for the 10 × 10, 5 × 5 and 3 × 3 layers: they had fewer peaks but still too many classification possibilities for clarity on an X, Y plot. Ultimately, the 2 × 2 self-organizing map processing layer provided the best results by forcing the data into a smaller number of categories. Figure 23a shows the two-dimensional X, Y plot using a 2 × 2 kohonen processing layer. Neglecting the data associated with final failure of the test object, the resulting X, Y plot for the 2 × 2 self-organizing map revealed five distinct, linearly separable, 54 failure mechanism 168 Acoustic Emission Testing F IGURE 21. A kohonen self-organizing map neural network.
Image of page 27

Subscribe to view the full document.

Image of page 28

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern