Abstract: In order to master the flatness characteristics of strips with different widths for ultra-wide tandem cold rolling mills, taking the sufficient flatness detection data of 1900 mm strips from 2180 mm cold rolling mills as a research object and considering the idea of big data and the cluster analysis method of data mining, this article proposed a cluster algorithm based on density and grid, applied this cluster algorithm to the analysis of flatness detection data under several typical strip widths, and then obtained the flatness characteristics of strips with different widths. A piecewise polynomial function was introduced to describe the strip flatness characteristics, and the analysis results of polynomial coefficients for strips with different widths were gotten. The proposed cluster algorithm based on density and grid and the piecewise function analysis method can be applied to analyze plenty of flatness detection data quickly and accurately, and the flatness defect characteristics and parameterized expression can be obtained, which will be a data basis of roll contour improvement and strip flatness control strategy optimization for cold rolling mills, especially ultra-wide cold rolling mills.