Research on acoustic temperature field hotspot tracking based on RBFNN adaptive grid

Acoustic tomography (AT) is an effective non-contact temperature distribution visualization technique. Most existing acoustic temperature measurement reconstruction algorithms are developed in relatively uniform temperature distribution environments, with less consideration given to acoustic reconstruction of temperature fields with extreme centre concentration and hotspot tracking problem. In order to improve the ability of acoustic tomography imaging to track temperature field hotspots, this paper established an acoustic temperature field reconstruction model based on an improved radial basis function neural network using AT technology. In addition, we proposed a grid partitioning and point layout scheme based on temperature gradient adaptation. When establishing the neural network model using this method, a polynomial was added as a correction term in the hidden layer to establish an improved radial basis function neural network (RBFNN) acoustic temperature field reconstruction model. Based on temperature gradient information without increasing computational complexity, Volonoi polygon embedding rule is applied to adaptively mesh the measurement area, complete the radial basis function centre point layout, reconstruct the temperature field, and track hotspots. Through numerical simulation, this method can effectively visualize the temperature field information when hotspots move to different positions. We reconstructed the temperature field hotspots on the self-developed acoustic temperature measurement experimental system and successfully detected the imaging of hotspots when the heater was moved by 0.25 meters. Simulation and experiments have shown that this acoustic method can effectively track hotspots.