A hybrid machine learning approach for designing a new XY compliant mechanism to maximize fatigue life

An XY-compliant mechanism is a two-degree of freedom positioner which is considered to have some distinguishing characteristics in ultra-precision technology. To ensure a long working life, it is necessary to study the fatigue life, however, in the related works of the XY-compliant mechanism, no research into design optimization for maximum fatigue life has been conducted so far. Moreover, the sample size of fatigue data is often inadequate to develop a surrogate model. This paper pioneers a new method for maximizing the fatigue life of the proposed mechanism, in the context of small-size fatigue life samples. Particularly, the Synthetic Minority Oversampling Technique is utilized to enlarge the sample size. A machine learning technique is then applied to create the surrogate model, in which fatigue life is considered the output. The Hunger Games Search is then used to maximize the output. In this study, Steel A36 and AL 6061T6 materials are utilized for the mechanism. The numerical results of two case studies show that using Synthetic Minority Oversampling Technique can predict appropriately the fatigue life, compared to using the original data. Specifically, after taking the natural logarithm of fatigue life, the average mean square error when using simulated data is \(1.8\%\) $1.8\%$ and \(36.8\%\) $36.8\%$ better than when using original data, for Case 1 and Case 2, respectively. The optimal fatigue life found by the proposed method compared with the baseline is about \(350\%\) $350\%$ for Case 1 and \(1250\%\) $1250\%$ for Case 2, respectively. The optimal findings are also confirmed using ANSYS software, in which the errors of fatigue life are less than \(5\%\) $5\%$ .