1.进行质量缩放的仅仅针对模型的均匀性不好，极少的部分单元尺寸很小，这时候进行质量缩放，增大极小单元的质量，间接增大整个模型的最小时间步长，但是需要注意的是，如果这些极小单元恰恰是你最关心的部位，这个时候质量缩放没有意义了！ 建议不要进行质量缩放！

2.如果网格差异比较大，有一种不常用的方法叫做子循环的方法，但是用的很少，只需要在模型中加入 *control -subcycle就行了 ，但是据说这种方法会带来稳定性方面的问题，引述如下：

Daniel  showed that this algorithm is in fact not stable in a classical sense, in the absence of any energy dissipation. Narrow timestep ranges are unstable, due to the nonlinearity of switching between the whole model updated once per major cycle, and the small timestep zone updated in minor cycles. As the model size increases, these unstable timestep ranges become extremely narrow, such that unstable states are very unlikely to be encountered. This situation has been labelled ‘‘statistical stability’’. The Belytschko et al. algorithm also has a second problem of possible inaccuracy due to a lack of momentum conservation at a timestep interface. This can occur due to the large timestep update only sampling the state at neighbouring small timestep nodes once per major cycle. 

我没有进行这方面尝试，对此也没有太多的发言权

3.还有一种方法，叫做子模型方法，这种方法也很少见有人使用，参考

http://www.rootfea.com/ShareView.Asp?ID=461

由于工作中自循环和子模型都没有涉及，不能提供具体的帮助，你可以问下其它的邻友是否有相关的经验