A Bayesian regularization-backpropagation neural network model for peeling computations

ABSTRACT

A Bayesian regularization-backpropagation neural network (BR-BPNN) model is employed to predict some aspects of the gecko spatula peeling, viz. the variation of the maximum normal and tangential pull-off forces and the resultant force angle at detachment with the peeling angle. $K$-fold cross validation is used to improve the effectiveness of the model. The input data is taken from finite element (FE) peeling results. The neural network is trained with $75\mathrm{\%}$ of the FE dataset. The remaining $25\mathrm{\%}$ are utilized to predict the peeling behavior. The training performance is evaluated for every change in the number of hidden layer neurons to determine the optimal network structure. The relative error is calculated to draw a clear comparison between predicted and FE results. It is shown that the BR-BPNN model in conjunction with the $k$-fold technique has significant potential to estimate the peeling behavior.

KEYWORDS:

• Machine learning
• adhesion
• peeling
• artificial neural networks
• bayesian regularization

Acknowledgements

The authors gratefully acknowledge the support from SERB, DST, under projects SB/FTP/ ETA-0008/2014 and IMP/2019/000276.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

¹ The nanoscale spatulae in geckos are very thin structures (approximately $5-10$ nm thick) with a width of around $200$ nm that can be modeled effectively as a thin strip.^{[14,15,38,39]}

² The correlation between the split and the predicted indices can be found in Table 2

Additional information

Funding

This work was supported by the Science and Engineering Research Board [IMP/2019/000276,SB/FTP/ETA- 0008/2014].