Process Analysis and Performance Assessment for Sheet Forming Processes
Singular Spectrum Analysis , controller performance assessment , minimum-variance , sheet forming process
Process monitoring and controller performance assessment are essential tools for ensuring that manufacturing processes operate safely, predictably, meet quality targets and operate profitably. Efficient techniques for process analysis and controller assessment facilitates are important for identifying areas for process and control improvement. Sheet- and film-forming processes pose special challenges for these techniques, because they often have periods of non-steady-state operation, and can exhibit both spatial and temporal variations. Existing process evaluation and analysis methodologies are oriented primarily for processes under steady operation, and focus on temporal variations. New tools are proposed to address limitations on the application of minimum-variance-based controller performance assessment to metal rolling processes. Extensions are proposed which address: 1) non-constant deadtime that arises from changes in rolling speed during startup, steady operation and wind down in the rolling process; 2) constraints on control actions; and 3) different sampling intervals for the manipulated variable input and the process output. The efficacy of the proposed extensions are demonstrated using an aluminum rolling mill and simulation examples. Singular Spectrum Analysis (SSA) is a promising technique for analyzing time series that decomposes data into a number of interpretable frequency components. New filtering and spectral interpretations of SSA are proposed in this work, including a modification to the computational procedure that produces a filter with zero-phase lag. Links between SSA are made to other signal processing and time series techniques. The potential for SSA in analyzing chemical manufacturing processes is demonstrated using an extended analysis of a two-tank process under periodic operation. An SSA-based approach is proposed for computing minimum-variance-based controller performance assessment, and as illustrated using an example. This technique has potential for providing more detailed diagnosis of elements causing poor controller performance. The effectiveness of a recently proposed two-dimensional SSA (2D-SSA) algorithm is investigated for the analysis of two-dimensional problems that frequently arise in sheet forming processes. The use and interpretation of this algorithm is demonstrated using two simple examples of rolling processes with known defects in the roll. 2D spectra are computed using the 2D-SSA algorithm, and are interpreted.