Ridge Identification for Hot Rolled Strip

摘要

Ridges are regions of comparatively thicker strip, typically 5 0 to 3 00 mm across the strip width but often extending over the full length of the coil. Even ridges of small height can cause downstream shape problems, particularly during cold rolling. It is therefore necessary for the strip thickness profile to be continuously monitored and the presence of any ridges noted promptly so that corrective action can be taken. Due to the ridge's small height, often less than 10 micrometers on hot-rolled strip, it can at times be difficult to differentiate the ridge from background measurement noise. This paper describes a computer-based real-time ridge identification system for hot strip rolling. From the raw measured data, the thickness profile is expressed as a series of Chebyshev polynomials. Once in this form accurate spatial filtering can be performed. By removing the high order polynomial terms, the high frequency measurement noise is reduced. In turn the low order terms include an accurate measure of the mean strip thickness, wedge and parabolic component of the strip profile. If these terms are also removed, it is shown that the remaining profile is one in which the ridges can be readily identified. Chebyshev polynomials were chosen as the basis functions for several reasons: to avoid the large amplitude oscillations common with many other filtering methods, the coefficients for the Chebyshev polynomials can be calculated quickly on standard desktop computers in the order of several milliseconds, and their calculation procedure is more robustthan other least-squares methods. Using this method the ridges are automatically detected from the measured thickness profile data and parameterised in terms of ridge height, ridge width and position across the strip width. Comparison with the presently used ridge detection method indicates a marked improvement in the ridge detection rate together with a reduction in the false alarms.