针对敏度过滤法拓扑优化结果存在边界扩散现象和过滤半径过大时图形过度磨平问题,提出一种新的敏度修正方法,将中心单元周围8个单元根据位置的不同分为两类,采用不同的加权系数,保持中心单元原目标函数敏度值的权重固定,三类单元加权求和。利用经典柔度最小化算例验证新的敏度修正方法的有效性。研究结果表明,新的敏度修正方法可有效消除棋盘格现象,体现较好的网格无关性,拓扑优化结果边界清晰,优化效果优于敏度过滤法。
Topology optimization results under sensitivity filtering method exists topology graph boundary diffusion, and the topology graph will be over-polished under the large filter radius. In order to solve the flaws of sensitivity filtering method , a sensitivity modifying method is proposed as a efficient approach, considered 8 elements around the center element, divided those elements into two categories and used different weighting coefficients.' Used constant weighting coefficient for center element, modified the center element objective function sensitivity by summing those objective function sensitivity in accordance with different weighting coefficients. The optimization effect of the new method is illustrated with classical minimize compliance problems. The experimental results show that the new sensitivity modifying method eliminated checkerboard patter and mesh independence, inhibited boundary diffusion. The new method have a better result on minimizing compliance.