Kinematic Prediction of a Bioinspired Foldable Flapping Wing Mechanism Using Artificial Neural Network

The presented research delves into the prediction of the kinematics for a bioinspired foldable flapping wing mechanism using an Artificial Neural Network. The investigation commences with the development of the kinematic parameterization for a foldable flapping wing UAV. Utilizing the established mathematical method of a foldable flapping wing mechanism, all datasets are generated to predict output parameters of links 4,7 and 8 (angular displacement) by modifying input parameters (fixed link length-crank radius ratio, angular displacement of crank, gear ratio, link-length ratio of special type of four bar mechanism) using MATLAB and verified by MSC ADAMS. The paper shows the implementation of a two-layer feed-forward neural network (a type of Artificial Neural Network), a subset of Machine Learning is employed to scrutinize the intricate relationships within the dataset using two algorithms under various geometrical conditions. The paper presents the comparison of the two algorithms i.e. (a) Levenberg–Marquardt algorithm and, (b) Bayesian Regularization algorithm of a two-layer feed-forward network. Out of this, the efficient algorithm is selected based on the performance parameters. This presented systematic approach needs no extensive experimental testing and support to make decisions based on predictions efficiently for all kinematic parameters of bioinspired foldable-flapping wing UAVs. This research may also be good for dynamic analysis to predict dynamic parameters for various mechanical systems.