中文摘要：

针对红外瓦斯传感器检测瓦斯浓度时易受到温度、压力等物理环境的影响而存在检测浓度值不准确的问题,提出一种结合分化距离与K-means算法方法找出瓦斯浓度数据集中的噪声点。该方法混合了分化距离和K-means算法的优点。对实验设计的传感器采集的瓦斯浓度数据集进行距离分化;运用Kmeans算法与孤立度系数结合法找出采集数据集中的噪声点并直接丢弃噪声点;通过均值法计算瓦斯浓度。实验结果表明：测量误差小于1%,精度高。

英文摘要：

Aiming at problem that infrared gas sensor is easy to be affected by temperature,pressure and other physical environment,which most likely bring noise points in gas concentration data set,always gives inaccurate results,propose a method which combines advantages of both differentiation distance and K-means algorithm,the method blended advantages of differentiation distance and K-means algorithm. Distance differentiation of gas concentration data set aquired by designed sensor is carried out; use K-means algorithm combined with isolation degree coefficient to find out noise points and discard them directly; mean value method is chosen to calculate gas concentration. Experimental results show that measurement error is less than 1 %,and precision is high.