DEVELOPMENT OF PREDICTIVE MODEL FOR MONITORING CONDITION OF HOT STRIPMILL

In the highly competitive steel industry, British Steel Strip Products (BSSP) has continually to focus on increased performance, product qua lity, and efficiency to maintain its market share and keep its custome rs. The hot strip mill (HSM) has traditionally been an area of some co ncern to BSSP owing to unscheduled mill breakdowns causing a loss of p roduction time and an associated reduction in product quality. BSSP ha s set up condition monitoring programmes to tackle some of these probl ems and is currently investing a great deal of money and resources to build on these initial successes. In this paper a review of BSSP's cur rent condition monitoring activities is overviewed. A prediction model is described which is intended to help improve and complement BSSP's main condition monitoring programme. The model was initially developed using artificial data, which mimic typical condition monitoring failu re data obtained from the Port Talbot HSM. It assumes that the failure pattern can be split into two phases, stable and unstable, which can be distinguished between by the use of a statistical process control m ethod. Depending on the progress of the failure, one of two models is used to predict the remaining machine life. The first is based on a re liability model, while the second uses a novel combination of reliabil ity and condition monitoring measurements. A series of failure case st udies based on actual HSM failures is used to test the model's predict ion performance. The applicability of the model to predict the useful life of a machine and optimise the time to repair/replace and a potent ial cost modelling strategy are discussed. (C) 1998 The Institute of M aterials.