针对车身检测数据样本量较小、某时域范围或某类型的测量数据呈现非正态分布的特点,提出了一种基于距离测度的车身检测数据多异常值排序检出算法,并推导了区分由生产过程突变造成的异常值和由测量粗大误差造成的异常值的判断方法.以某车型的间隙与面差检测数据为例,验证了本方法的有效性.
In view of the features of vehicle body measured data,i.e.small sample and non-normal distribution in certain time or of certain type,a distance-based multi-outliers sorting/detection algorithm for vehicle body measurement is proposed,and a judgment criterion is derived for distinguishing the outliers caused by sudden changes in production process from the outliers caused by gross error of measurement.The method is validated with a real example of gap and flush measurement of a vehicle.