shape identification using distributed sensing technology
Mission critical structures , such as plane wings, turbine blades, bridges, cranes, and space booms, will require techniques for monitoring their health to ensure mission success. In certain applications, such as morphing air foil, real-time shape sensing techniques could be used to control the air foil shape. To support this goal, research on shape identification and mode shape estimation is underway by various researchers, and this research is gaining traction recently due to the parallel technology development of distributed sensing technologies, such as Fiber-Bragg optical fiber strain/temperature sensing device.
As distributed sensing technology revolutionizes the measurement method, it equally revolutionizes on how the measured data can be utilized for various applications. SIMAT's research interest is how to use distributed sensing technology for shape identification, mode shape analysis, and component life prediction using theories widely used in point based sensing technology.
Recently, SIMAT developed an approach to compute beam deflections from experimental data using least-squares curve-fitting and integration method. This method is different from Kho's method wherein beam deflections are computed using closed form solutions. A comparative result between SIMAT's and Kho's method is shown in the figure (see to the right). The strain data is obtained from a research paper "Displacement theories for in-flight deformed shape predictions of aerospace structure," W L Kho, W L Williams, V T Tran, 2007, NASA Dryden Flight Research Center. Deflection obtained from Kho's method and SIMAT's curve-fitting approach are in close agreement.
Research to estimate mode shapes using a combination of finite element model and measured strain data is in progress.
Note: The deviation between Kho and SIMAT's method at station 6 and 7 is possibly due to inference error from figure 11 in the reference research paper.
Manufacturing part quality verification using AI
Verification of manufactured parts using conventional techniques, such as human or computer vision, leads to either slow throughput or inaccurate qualification. Identifying a part to be acceptable when it should be rejected and vice-versa is common in high-volume production due to human error. With computer vision based verification systems - one without AI algorithms - can identify only type of part. Change of part in the production line warrants a change in the inspection system, which is time-consuming. The traction in AI and its potential benefit in manufacturing industry, specifically in part quality verification, is what we are researching to address the limitations mentioned above through the development of deep learning algorithms. To support this venture, SIMAT has joined NVIDIA's inception program. Specific details about this research program will be published soon.