A Hybrid Bayesian-Genetic Approach for Micro Gear Tolerance Optimization

In precision engineering, particularly in microgear production, stringent tolerance levels are essential for ensuring optimal functionality. Traditional tolerance design models in this field focus primarily on product value, often overlooking aspects such as product robustness, which presents challenges in maintaining a balance between manufacturing precision and cost efficiency. This study proposes a hybrid optimization method that combines the predictive strengths of Bayesian techniques with the comprehensive search capabilities of genetic algorithms. The proposed methodology employed integrates statistical analysis with algorithmic modeling. Bayesian methods are utilized for forecasting and adjusting tolerance levels, using historical data and probabilistic models to enhance manufacturing accuracy. Concurrently, genetic algorithms are being applied to explore a range of design parameters, aiming to identify optimal tolerance settings. Combining these methods allows for an extensive exploration of potential solutions, seeking to achieve an equilibrium between precision and cost. Experimental results on a set of micro gear design cases demonstrate that the proposed BayesianGenetic Approach approach significantly outperforms the standalone Bayesian Optimization method and achieves competitive results compared to the standalone Genetic Algorithm method.