Tribological Behaviors and Coefficient of Friction Estimation of Textured Leaded Brass/C45 Friction Pairs Utilizing Time-Series Prediction Method of CPO-BiLSTM

Abstract Leaded brass/C45 steel tribo-pairs have been widely used in the aerospace industry, however, failure often occurs due to severe operating conditions and fatigue. Surface texturing is a promising technique for enhancing tribological properties, thus targeted experiment is essential to conduct for determining reasonable structural parameters. Moreover, it is costly and time-consuming to study the tribological behavior, and the experimental data are time-series that should be further utilized. Therefore, the tribological behaviors and wear mechanism of textured leaded brass/C45 steel friction pairs lubricated by 10# aviation hydraulic oil with six texture sizes and two rotating speeds were researched by wear test, with the aim of optimizing the tribological characteristics of the textured friction pairs. Furthermore, a time-series bi-directional long short-term memory (BiLSTM) prediction model optimized by the crested porcupine optimizer (CPO) algorithm was proposed to predict the friction coefficients. Experimental results have shown that the cylindrical dimple texture demonstrated better tribological behaviors than that of the reticular texture because abrasion was aggravated by the flow of wear debris through the grooves. As the texture size increased, the friction coefficient and wear-rate first decreased and then increased, because the contact stresses also increased with the size, leading to more rapid damage to the texture. Additionally, the proposed CPO-BiLSTM algorithm accurately predicted the friction coefficients by automatically searching for the optimal parameters of the BiLSTM model within a given parameter range, which provided a valuable direction for life prediction of textured tribo-pairs.