目的 针对在与国外某汽车材料实验室进行冷轧薄钢板表面粗糙度比对试验中,没有统一评价判据的问题,对交流中提及的相关性和一致性进行系统研究。 方法 对评价方法遴选中涉及的数据,进行相关系数检验和配对 t 检验。 结果 经汽车板表面粗糙度比对试验数据验证,相关系数检验只适合评价两组试验数据是否具有线性相关关系;配对 t 检验,则适用于评价两组试验数据的均值之间是否具有一致性。 通过应用实例,介绍了一种新开发的,在同一界面中既能显示配对 t 检验的定量评价结论,又能通过比对试验数据散点图几何中心点和坐标原点连线与 45°理想位置线夹角的数值小于限制夹角为±1°的标准,直观显示比对试验的评价结果,编写完整结论的比对试验评价软件。 ±1°的评价标准是由大量的汽车板表面粗糙度比对试验数据统计确定的。 结论 由于汽车板表面麻点分布不均匀,所以在汽车板表面粗糙度比对试验中,只能采用一致性评价方法进行判断评定。

Objective Aiming at the problem of not having a unified evaluating criterion for the contrast test during the comparison with a foreign auto material lab on the surface roughness of cold-rolled steel sheets, systematic study was carried out on the correlation and consistency mentioned in the communication. Methods This paper elaborated the correlation coefficient test and the paired t-test both concerned in the selection of evaluation methodologies. Results By the contrast test data verification of the auto sheet surface roughness, the correlation coefficient test was only suitable for evaluating whether the two groups of test data was linearly correlated; the paired t-test was suitable to evaluate whether there was consistency between the averages of the two groups of test data. Based on application examples, a newly developed evaluation software of the contrast test was introduced. On the same interface, the software could not only display the quantitative evaluation conclusion of the paired t-test, but also visually display the evaluation result on the scattergram of the contrast test data via the value that was smaller than the limiting angle of ±1° standard of the angle formed by the line connecting the geometric center point to the origin of coordinate and the 45° ideal location line, and then draw a complete conclusion. ±1° standard was determined by a large number of statistical surface roughness contrast test data for the auto sheet. Conclusion Due to the uneven distribution of surface pits on the auto steel sheets, only the evaluation method for consistency could be used in the contrast test of auto sheet surface roughness.