This work is dedicated to the solution filtering technique for performing direct and large-eddy simulation. It is shown that this approach is equivalent to the use of spectral viscosity as a possible ersatz of subgrid-scale modelling. In the framework of finite-difference schemes, the filter operator can be designed to ensure time consistency while easily controlling the level and scale selectivity of the dissipation thus introduced. Then, a new family of filter schemes is developed in order to represent both the molecular and artificial dissipations. The resulting viscous filter operator is straightforward to implement through a simple modification of its coefficients that depend on the molecular/artificial viscosity and the time step. A definitive advantage in terms of computational efficiency is obtained for computational configurations where the time step is restricted by the Fourier condition, as a simple alternative to implicit time integration of the viscous term. Numerical tests clearly show that this viscous filtering method is flexible, accurate and numerically stable at large Fourier number.

这项工作专门用于执行直接和大涡流仿真的解决方案过滤技术。结果表明，这种方法等效于使用光谱粘度作为亚网格规模建模的可能结果。在有限差分方案的框架内，可以设计滤波器算子以确保时间一致性，同时轻松控制由此引入的耗散的电平和标度选择性。然后，开发了一个新的滤波器方案系列，以表示分子耗散和人工耗散。通过简单地修改其系数（取决于分子/人工粘度和时间步长），可以轻松实现最终的粘性过滤器运算符。对于时间步长受限于傅立叶条件的计算配置，它是计算效率方面的绝对优势，可以作为粘性项隐式时间积分的简单替代方案。数值试验清楚地表明，这种粘性过滤方法在大傅立叶数下是灵活，准确和数值稳定的。