Optimal Design of a Cold Spray Nozzle for Inner Wall Coating Fabrication by Combining CFD Simulation and Neural Networks

Yuxian Meng, Hiroki Saito, Chrystelle Bernard, Yuji Ichikawa, Kazuhiro Ogawa

研究成果: ジャーナルへの寄稿学術論文査読

抄録

Recently, the low-pressure cold spray (LPCS) technique has been used to fabricate superhydrophobic polymer coatings on metallic substrates, suggesting a significant potential in engineering applications. This study aims to design a spiral LPCS nozzle to coat the pipe’s inner wall with superhydrophobic polymer. The design goal is to achieve the maximum particle velocity in a confined (limited) space, assuming that the powder can enter the feeding tube through the Venturi effect. Achieving these two goals simultaneously using only computational fluid dynamics (CFD) simulation is challenging. Therefore, the CFD simulation was combined with the neural network (NN) method to design the new spiral nozzle. During training, the effects of the NN models and algorithms were investigated. The results showed that the feedforwardnet model combined with the trainbr or trainlm algorithm (from MATLAB 2016b software), presented a minimal error for particle velocity or gas flux prediction, respectively. The trained NN correlates the nozzle parameters (i.e., mean coil diameter, spring lift angle, and expansion ratio) and its performances (i.e., particle velocity and gas flux in the powder feeding tube). As a result, the optimal spiral nozzle was determined based on the design goal of maximum particle velocity and suitable gas flux in the powder feeding tube. Furthermore, the effect of each nozzle parameter on the particle velocity and gas flux in the powder feeding tube was analyzed. The cold spray experiment confirmed that the designed spiral nozzle could fabricate Perfluoroalkoxy alkane (PFA) coatings.

本文言語英語
ページ(範囲)3-16
ページ数14
ジャーナルJournal of Thermal Spray Technology
33
1
DOI
出版ステータス出版済み - 2024 2月

フィンガープリント

「Optimal Design of a Cold Spray Nozzle for Inner Wall Coating Fabrication by Combining CFD Simulation and Neural Networks」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル