<p>The process of high-precision grinding of ceramic rollers is characterised by a series of consecutive machining stages, including rough grinding, semi-finish grinding, finish grinding, and spark-out grinding. Each of these stages necessitates a distinct set of machining process parameters. The design of multi-pass grinding processes is typically dependent on empirical knowledge and complex modelling. In scenarios where multiple grinding process output targets are present, intricate interactions between grinding stages emerge, impeding the balance between surface quality and machining efficiency. In this paper, a multi-objective multi-process grinding process parameter optimisation method for the outer contour of silicon nitride (Si₃N₄) spherical roller is proposed. A prediction model for the evolution of surface roughness (<i>R</i>_<i>a</i>) and roundness error (<i>CE</i>) in the grinding process of Si₃N₄ spherical roller is established by Response Surface Methodology (RSM). To achieve the optimisation of the process parameters of Si₃N₄ spherical roller outer contour grinding, the Adaptive Strategy Particle Swarm Optimization (ASPSO) algorithm has been adopted. A comparison of the searching ability of ASPSO with that of the multi-objective particle swarm algorithm (MOPSO) demonstrates the superior searching accuracy of the former in dealing with complex optimisation problems. The efficacy of the proposed methodology is substantiated by grinding experiments, with the discrepancies between the predicted and experimental values of <i>R</i>_<i>a</i> and <i>CE</i> amounting to 11.53% and 9.53%, respectively. A comparison of the experimental results of empirical process parameters, the grinding efficiency is improved by 16.61% while maintaining high surface accuracy of the cylindrical surface.</p>