Machine learning modeling and multi objective optimization of artificial detrusor

To address the problem of obtaining optimal design parameters for existing artificial detrusors using single-objective optimization methods, this research proposed a machine learning-based artificial detrusor modeling and multi-objective optimization approach, which includes a thorough process from modeling to optimization. Extreme learning machine was used to model artificial detrusors in the suggested approach, and in order to increase modeling accuracy, a multi-strategy modified crayfish optimization algorithm for tuning the extreme learning machine's parameters was put forth in this research. The multi-objective grey wolf optimization algorithm was utilized to optimize the artificial detrusor based on the model. In order to validate the suggested approach, an artificial detrusor driven by a shape memory spring was finally built as an experimental platform. The results show that the improved crayfish optimization algorithm proposed in this paper can effectively avoid the defects of the original algorithm, and its optimization performance and convergence ability are better than the comparison algorithm. With a root mean square error of 1.51E-02, a coefficient of determination of 9.81E-01, a mean absolute error of 1.32E-02, and a mean absolute percentage error of 1.66E-01, the established artificial detrusor model predicts the shape memory spring-driven artificial detrusor's emptying rate. It also predicts the temperature increment with a root mean square error of 8.47E-01, a coefficient of determination of 9.81E-01, a mean absolute error of 5.81E-01, and a mean absolute percentage error of 7.23E-02. These predictions are superior to the comparison prediction model, indicating good predictive performance and stability. Additionally, the established artificial detrusor model also demonstrates outstanding performance in uncertainty and reliability analysis, thereby further confirming its superior comprehensive performance. The optimized artificial detrusor's computed values of emptying rate and temperature increment, as well as its experimental measurement values, have errors of 7.8% and 11.8%, respectively, which satisfy engineering design specifications. The artificial detrusor optimized by our proposed method exhibits significant performance enhancements over existing designs. Specifically, the optimized detrusor achieves an approximately 20% increase in emptying rate and a 62% reduction in temperature increment, successfully balancing urinary efficiency with mitigated risks of thermal tissue injury.