Optimization of training sets for neural-net processing of characteristic patterns from vibrating solids

Decker, Arthur J.

doi:10.1117/12.449379

Cited by 3 publications

(7 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…14 A later study showed that the coding, conditioning and transformations of the input patterns affected significantly the magnitude of the minimum detectable change in the vibration amplitude distribution. 13 The results of these studies have proven to be effective for both model and experimental data. The design rules listed next were developed for feed-forward nets that contain one hidden layer.…”

Section: Design Of Neural Nets and Training Sets For Damage Detectionmentioning

confidence: 91%

“…It was necessary to amplify the difference in displacement distributions, before the net would learn the difference. A data transformation technique called folding 13 changed that result significantly. Following folding, the crack could be de-amplified, and the net would learn to detect it.…”

Section: Design Of Neural Nets and Training Sets For Damage Detectionmentioning

confidence: 99%

“…13 Early experiments with the raw absolute value patterns represented by equation (1) had difficulty distinguishing the model-generated cracked and uncracked cantilever modes. It was necessary to amplify the difference in displacement distributions, before the net would learn the difference.…”

Section: Design Of Neural Nets and Training Sets For Damage Detectionmentioning

confidence: 99%

“…The entire pattern can be compared somehow with a pattern from the undamaged component [11][12][13] or even with a design-model-generated pattern. [14][15][16] This approach does not depend on the presence of dramatic local structural effects such as cracks.…”

Section: Whole-field Approachesmentioning

confidence: 99%

See 3 more Smart Citations

Damage Detection Using Holography and Interferometry

Decker

2003

Optical Metrology for Fluids, Combustion and Solids

Self Cite

View full text Add to dashboard Cite

Section: Design Of Neural Nets and Training Sets For Damage Detectionmentioning

confidence: 91%

Section: Design Of Neural Nets and Training Sets For Damage Detectionmentioning

confidence: 99%

Section: Design Of Neural Nets and Training Sets For Damage Detectionmentioning

confidence: 99%

Section: Whole-field Approachesmentioning

confidence: 99%

See 2 more Smart Citations

Damage Detection Using Holography and Interferometry

Decker

2003

Optical Metrology for Fluids, Combustion and Solids

Self Cite

View full text Add to dashboard Cite

“…This work is admittedly critical for selecting the correct beam conditioning, but is in progress by other researchers, and will not be discussed in any detail. Finally artificial neural networks have been used successfully for pattern-based process control 6,7 and fringe-pattern interpretation 8,9 for more than a decade. The possibility of using neural nets to interpret the post-scattering interference patterns and change the input interference patterns will be discussed briefly.…”

Section: Introductionmentioning

confidence: 99%

Interferometer control of optical tweezers

Decker

2002

Interferometry XI: Applications

Self Cite

View full text Add to dashboard Cite

This paper discusses progress in using spatial light modulators and interferometry to control the beam profile of an optical tweezers. The approach being developed is to use a spatial light modulator (SLM) to control the phase profile of the tweezers beam and to use a combination of the SLM and interferometry to control the intensity profile. The objective is to perform fine and calculable control of the moments and forces on a tip or tool to be used to manipulate and interrogate nanostructures. The performance of the SLM in generating multiple and independently controllable tweezers beams is also reported. Concurrent supporting research projects are mentioned and include tweezers beam scattering and neural-net processing of the interference patterns for control of the tweezers beams.

show abstract

Damage Detection Based on ESPI and SVM

Liu

Meng

Huang

2006

KEM

View full text Add to dashboard Cite

An approach of damage detection based on ESPI and SVM is proposed. ESPI (Electronic Speckle Pattern Interferometry) is a non-contacting measuring method, which can measure the small static and dynamitic surface deformations and reveal the flaws by looking for flaw-induced deformation anomalies. Support Vector Machine (SVM) is a machine learning algorithm based on statistical learning theory, and it has recently been established as a powerful tool for classification and regression problems. To develop the precision of processing the pattern fringe data, the SVM is introduced to process the patterns corrupted by the laser speckle effect. The SVM is trained with fringe patterns generated from a finite element model and a simple model of the laser speckle effect. The output pattern is obtained to flag whether the damage exists or not. The trained SVM is tested for robustness with model generated test patterns of a flat plate. The results show that this approach is a promising and effective for damage detection.

show abstract

Optimization of training sets for neural-net processing of characteristic patterns from vibrating solids

Cited by 3 publications

References 0 publications

Damage Detection Using Holography and Interferometry

Damage Detection Using Holography and Interferometry

Interferometer control of optical tweezers

Damage Detection Based on ESPI and SVM

Contact Info

Product

Resources

About