Abstract: An improved sine and cosine pathfinder algorithm with refracted opposition-based learning is proposed for the function optimization problem to address the problems of slow convergence, low accuracy of the pathfinder algorithm, and the tendency to fall into local optimality. First, the population is initialized by a refracted opposition-based learning strategy. The combination of refraction and inverse principle is used to make the initial solution closer to the optimal solution position, and the high-quality population positioning can provide a good basis for the strategy execution in the iteration period. Second, an improved sine and cosine individual position update approach is introduced in the pathfinder position update phase, which replaces the linear step search factor in the original update equation to generate new generations of pathfinder individuals in a non-regular pattern, thus reducing the probability of individuals ignoring the optimal solution. An adaptive weight is also proposed to be added to the original update formula to balance the global search and local exploitation ability of the algorithm with the sine and cosine functions. Finally, the proposed algorithm is applied to twelve classical benchmark test functions and ten CEC2014 benchmark test functions with complex features for finding the optimal solution, and it is applied to pressure vessel design and three-rod truss design problems, and suitable evaluation indexes are selected to assess the performance of the algorithm. The experimental results show that the proposed algorithm has improved in terms of convergence speed, optimization-seeking accuracy, and local optimum avoidance, and the excellent engineering optimization performance also demonstrates the robustness of ours.