| 摘要 |
Process for identifying causes of defects in sheet- rolling mills, which makes it possible to determine, in probability terms, the causes which lead to anomalous operation in a semi-continuous rolling mills, and consequently to a defective sheet. The method requires only historical information on the process, and can itself be equipped with a self-explanatory model of the process itself based on neural networks or else use the existing one for regulating the plant. On the basis of the model, and using an evolving strategy with genetic alteration by mutation, solutions are sought from the point of view of the error. By proximity, having the historical information on the process available, it is possible to assign probabilities of occurrence. The process proves to be useful for predicting causes of defects expressed as deviations in widths (out of tolerance), temperature deviations, etc., both analogue and digital, which convey a particular state of the mill.
|
