目的 提升氮化硅陶瓷加工质量和效率，提高粗糙度模型预测精度。方法 提出塑性与塑-脆性去除转变临界切深hc1和塑-脆性与脆性转变临界切深hc2，然后对原有模型进行修正，并引入塑性去除粗糙度修正系数φ1、τ1和塑-脆性去除粗糙度修正系数φ2、τ2，建立基于不同去除方式的粗糙度Ra预测模型，后通过磨削实验对系数进行求解，并得出磨削参数对粗糙度和表面形貌的影响。结果 塑性去除粗糙度修正系数φ1=5.872×10–6、τ1=0.1094，塑-脆性去除粗糙度修正系数φ2=1.299×10–5、τ2=0.1582。砂轮线速度vs由30 m/s增大到50 m/s，粗糙度Ra由0.366 μm减小到0.266 μm，去除方式由脆性断裂向塑性变形转变，表面质量变好。磨削深度ap由5 μm增大到45 μm，粗糙度Ra由0.252 μm增大到0.345 μm，去除方式由塑性变形向脆性断裂转变，表面质量变差。工件进给速度vw由1000 mm/min增大到9000 mm/min，粗糙度Ra由0.227 μm增大到0.572 μm，去除方式由塑性变形向脆性断裂转变，表面质量变差。模型预测值与实验值的相对误差δ在2.1%~8%之间。结论 在加工中应控制磨削深度和工件进给速度，适当提高砂轮线速度，以保证加工精度和效率。基于不同去除方式的粗糙度预测模型，可较为精准地预测实际加工情况。

The work aims to improve the processing quality and efficiency of silicon nitride ceramics, the prediction accuracy of the roughness model. The critical depth of plastic and plastic-brittle removal transition hc1, critical depth of plastic-brittle and brittle removal transition hc2 were proposed. Secondly, the original model was modified. The roughness correction coefficients of the plastic removal φ1 and τ1, the roughness correction coefficients of the plastic-brittle removal φ2 and τ2 were introduced. A roughness prediction model based on different removal methods was established. Finally, the coefficients were solved by grinding experiments. The effect of grinding parameters on roughness and surface topography were obtained. The results show that the roughness correction coefficients of the plastic removal φ1 =5.872×10–6, τ1=0.1094. The roughness correction coefficients of the plastic-brittle removal φ2 =1.299×10–5, τ2=0.1582. When the grinding wheel speed vs increases from 30 m/s to 50 m/s, the roughness Ra decreases from 0.366 μm to 0.266 μm. The removal method changes from brittle fracture to plastic deformation. And the surface quality becomes better. When grinding depth ap increases from 5 μm to 45 μm, the roughness Ra increases from 0.252 μm to 0.345 μm. The removal method changes from plastic deformation to brittle fracture. And the surface quality deteriorates. When workpiece feed rate vw increases from 1000 mm/min to 9000 mm/min, the roughness Ra increases from 0.227 μm to 0.572 μm. The removal method changes from plastic deformation to brittle fracture. And the surface quality deteriorates. The relative error between the model predictive value and the experimental value δ is 2.1%~8%. The conclusion is that the grinding depth and the workpiece feed speed should be controlled during the machining. And the grinding wheel speed should be properly increased to ensure the machining accuracy and efficiency. The roughness prediction model based on different removal methods can accurately predict actual machining conditions.