关键词: 磨料水射流, 钛合金, 神经网络, 粒子群算法, 轮廓

Abstract: Titanium alloy is widely used in the field of aerospace, but it is difficult to process and suffers the problem of low processing efficiency. Abrasive waterjet has many advantages such as cold working and high energy density, and starts to be an effective method for processing titanium alloy. In order to study the characteristics titanium alloy processed by abrasive waterjet, milling experiments using TC4 specimen with a single pass were designed and conducted by considering the influences of jet pressure, abrasive flow rate, stand-off distance, jet angle, and feed rate. Moreover, the maximum depth h_max, maximum width b_max, and half-depth width b_0.5 of the milled surface were used as the evaluation objectives. An empirical model for predicting these three parameters was established by dimensional analysis, a BP neural network prediction model was established according to the strong non-linear characteristics of milling factors affecting the profile, and BP-PSO model was proposed by introducing PSO algorithm to optimize the weight of the model globally. The results show that the average errors of the three models are all less than 10%. The prediction accuracy of the empirical model was higher than that of the BP network model but lower than the PSO-BP network model. Compared with the BP network model, the average errors of maximum depth h_max, maximum width b_max, and half-depth width b_0.5 of PSO-BP network model are reduced by 25.09%, 14.67%, and 9.66%, respectively. The optimized model significantly improves the accuracy for predicting the surface profile features of TC4 titanium alloy milled by abrasive waterjet, and also provides reasonable process parameters.

Key words: abrasive waterjet, titanium alloy, neural network, particle swarm algorithm, profile