Application of Neural Networks for Simultaneous Stereoscopic Tracking of Numerous Particles in Flow and Motion Measurements Measurements

Stereoscopic tracking technology has been developed for simultaneous tracking of numerous markers including particles and dots in three-dimensional space. It can offer an unprecedented capability for measuring three-dimensional (3-D) velocities in a fluid flow and displacement/deformation of a solid object. For fluid motion measurement, the flow can be seeded with small particles. For surface motion, the surface can be marked with dots while for deformation of a transparent object, it can be embedded with particles. In our effort for implementing the stereoscopic tracking technology, we employ neural networks for particle image overlap decomposition and simultaneous stereoscopic tracking of several thousands markers with high-resolution high-speed solid state cameras. One of the successful application examples is the measurement of parachute aerodynamics, for which the technique is termed stereoscopic tracking velocimetry (STV). By seeding the measurement volume with neutrally-buoyant helium soap bubbles as velocity markers, the parachute flow field can be visualized. The testing with a medium-scale model is conducted with high-speed high-resolution cameras. It is not feasible to provide a 3-D flow field of known velocity for assessing accuracy in 3-D flow measurement. For measurement accuracy and performance estimation, known velocity fields and corresponding stereoscopic particle images are thus obtained through 3-D numerical simulation of the aerodynamics flow around a parachute. Another successful application example is the deformation measurement of the brake system component of a heavy-duty truck, that is, slack adjuster. For the reliability evaluation of the measurement technology, numerical calculation for the stress analysis has been also performed for cross examination.