Acoustic measurement of velocity filed using improved radial basic function neural network. (March 2023)

• Record Type: Journal Article
• Authors: Kong, Qian; Lu, Ying; Jiang, Genshan; Liu, Yuechao
• Is Part Of: International journal of heat and mass transfer. Volume 202(2023)
• Publication Date: 2023-03
• Subjects: Acoustic method -- Velocity filed measurement -- Neural network -- Radial basic function -- Furnace; Heat -- Transmission -- Periodicals; Mass transfer -- Periodicals; Chaleur -- Transmission -- Périodiques; Transfert de masse -- Périodiques; Electronic journals; 621.4022
• DOI: 10.1016/j.ijheatmasstransfer.2022.123733
• Languages: English
• ISSNs: 0017-9310

Abstract

Highlights: A novel improved RBFNN model based on acoustic tomography is established to reconstruct the velocity field. The influence of various parameters in the neural network is studied to improve the reconstruction performance. The reconstruction numerical results obtained using the RBFNN method have higher accuracy, better real-time performance, and noise immunity as compared with other acoustic algorithms. The results of proposed acoustic measurements in lab prove the effectiveness in practical velocity field measurement. Abstract: In coal-fired boilers, the aerodynamic field directly affects the combustion conditions of the boiler and safety of the furnace. In this work, a new method based on acoustic tomography (AT) is proposed to measure the velocity field by using radial basis function neural network (RBFNN). First, we establish the proposed RBFNN model to reconstruct the velocity field and study the influence of various parameters including the number and arrangement of sensors, the shape parameters and the number of center points in RBF, and the hyper parameters in the neural network to improve the reconstruction performance. Second, a novel improved RBFNN acoustic algorithm is used to reconstruct different two-dimensional velocity fields based on numerical simulations. The reconstruction results obtained using the proposed algorithms have higher accuracy, better real-time performance, and noise immunity as compared with other classical acoustic algorithms. Third, the physical experiments are conducted in lab to further assess the proposed algorithm. The results exhibit that the proposed RBFNN acoustic method provided satisfactory consistency as compared to the traditional measurement approaches. This proves the effectiveness of the proposed acoustic method in practical velocity field measurement.