An improved support vector machine (ISVM) surrogate model is developed in this paper, which combines Bayesian optimization with grid search. By adaptively adjusting hyperparameters, the proposed model achieves higher prediction accuracy and stronger generalization, and is further applied to the reliability evaluation of mechanism motion precision. Firstly, two numerical case studies are conducted to compare the ISVM with the traditional support vector machine (TSVM), thereby demonstrating the advantages of the proposed model in terms of prediction accuracy. Secondly, a crank-slider mechanism with clearance is employed as a case study, where an ISVM surrogate model is developed to capture the mapping between multiple uncertain parameters and the maximum displacement error. By incorporating the stress-strength interference theory, a reliability model is further formulated under the conditions of multidimensional random variables. Finally, Monte Carlo simulation (MCS) is utilized to assess the failure probability and reliability level of the mechanism, followed by a systematic analysis of the reliability characteristics under different failure thresholds. The results indicate that the ISVM achieves higher prediction accuracy than TSVM, at the cost of a moderate increase in computational effort, providing solid theoretical support and practical guidance for reliability modeling and optimal design of complex mechanical systems under uncertainty.
Yujie Zhang (Wed,) studied this question.