Aiming attheproblemsoflow convergenceaccuracyofgazelleoptimization algorithm and easyto fallinto local optimum , an elite inverse learning and Cauchy perturbation-guided improved gazelle optimization algorithm (IGOA) is proposed. Firstly, the gazelle individuals areinitialized using an elite reverse learning strategy to improve the quality of the initialsolution and increase the population diversity. Secondly, atthe beginning ofthe algorithm iteration, the two- stage nonlinearinertia weights are used to guide the position updating method of the population, which improves the accuracy ofthe algorithm and balances the global and local searches of the algorithm. Finally, the survival rate-guided Cauchy perturbation strategyisintroduced into theposition updatingformulaofthepopulation in theexploration stageto improve theabilityofthealgorithm tojump outofthelocaloptimum. Experimentaltestsarecarried outon8 comparison algorithms using 12 benchmark test functions and Wilcoxon rank sum test, and the results show that the improved algorithm has higher optimization accuracy, faster convergence speed and ability to jump out of local optimum. The practicality and effectiveness ofIGOA are verified on two real engineering problems, namely, gear train and three-bar truss design.