Low frequency signals fig 13 of 100 khz were

This preview shows page 17 - 19 out of 47 pages.

Low frequency signals (Fig. 13) of 100 kHz were generated as a result of the weld puddle that formed and flowed. The amplitudes and energies of these signals were generally small, with rise times similar to the type 1 signal. In certain instances, very large, up to 90 dB, low frequency signals were produced and were thought to be caused by the welder’s bumping the panel with the welding rod. In certain instances, during the repair, abnormal weld arcs produced spikes in the recorded waveforms. These signals had large amplitudes and energies with small rise times and number of counts (Fig. 14). The power spectrum of these signals would essentially be a flat line beyond 1 MHz because of the dirac delta function nature of the spike. If the acoustic signals were as well behaved as these examples and as few in number, identification of hot cracking would be no problem. In reality, because of source arrival time overlap, discontinuity location, number of signals and other considerations, it is very difficult to sort each signal by hand. The solution is first to develop a neural network mode to classify each signal by first categorizing all of the signals 238 Acoustic Emission Testing F IGURE 13. Rubbing and weld flow signal: (a) amplitude; (b) frequency. Signal intensity (mV) Time (μs) 50 40 30 20 10 0 –10 –20 –30 –40 –50 Signal intensity (arbitrary unit) Frequency (kHz) 0 199 398 598 797 1000 (a) (b) 0 32 64 96 128 160 192 224 256 10.000 1.000 0.100 0.010 0.001 F IGURE 14. Spark noise signal: (a) amplitude; (b) frequency. Signal intensity (V) Time (μs) 4 2 0 –2 –4 –6 –8 –10 0 32 64 96 128 160 192 224 256 Signal intensity (arbitrary unit) Frequency (kHz) 100.0 10.0 1.0 0.1 0.01 0.001 0 199 398 598 797 1000 (a) (b)
Image of page 17

Subscribe to view the full document.

recorded during the weld into a particular signal class and then to identify the mechanism that creates each signal. Neural Network Data Analysis Self-organizing map artificial neural networks were developed using software to map features of the parametric data, waveform data and parametric waveform data. The self-organizing map is a three-layered network with an input and a kohonen classification layer, plus an x,y coordinate output layer (Fig. 15). In the input layer, data are ordered and presented to the network as an N -dimensional input vector. As an example, N could represent the signal rise time, counts, energy and amplitude from the parametric acoustic emission data set. Fully connected to the input layer, the kohonen layer provides a two-dimensional grid of neurons through which unsupervised learning takes place. The result of training the self-organizing map network is a topological map of the multidimensional input vector space, where order is preserved through a grouping of input data with similar features. The self-organizing map performs the data clustering through a minimization of the euclidean distance between the kohonen layer weights and each input data vector.
Image of page 18
Image of page 19

{[ 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