粗集理论（Rough Sets Theory，RST）在许多领域已得到成功的应用，它在处理不完全、不精确、不确定性智能信息问题中表现出优秀品质。然而在标准RST理论中存在有局限性，比如对等价关系的定义使得知识约简相对复杂化。论文提出了基于RST的容错粗集的概念，给出了容错粗集上的性质，使得在进行知识约简时无需进行量化。从而大大提高了知识发现的效率。 Rough Sets Theory (RST) has been successfully applied in many fields. It shows excellent qualities in dealing with incomplete, inaccurate and uncertain intelligent information problems. However, there are limitations in the standard RST theory. For example, the definition of equivalence relation makes the knowledge reduction relatively complicated. In this paper, the concept of fault-tolerant rough sets based on RST is proposed, the properties of fault-tolerant rough sets are given, which makes it unnecessary to quantify the knowledge reduction. Thus greatly improving the efficiency of knowledge discovery.