Bees Swarm Optimization Based Prediction of Machined Fatigue Failure

Artificial intelligence approach is now able to anticipate fatigue failure of machined components based on elementary key numbers without the need for professional knowledge of evaluation processes. In this paper, we develop a meta-heuristic approach for the prediction of defects using Bees Swarm Optimisation (BSO) in machined components. The study mainly predicts the fatigue failure by training the BSO experimentally with the training datasets. The three most essential input variables are the load stress, the hardness of the material, and the size of the fatal flaw (KD). In the context of fatigue testing, the only outcomes that may be determined are failure or non-failure. In this investigation, a single data set is generated for each specimen that is examined. When a full fracture has occurred at a given stress level, fracture surface analysis estimates the size of the deadly flaw that has formed. The architecture and hyperparameters of the neural network are cross-validated using a K-fold technique in order to get the highest possible prediction accuracy. After all is said and done, a cautious mean fatigue failure prediction accuracy of 92% is achieved.