Fuzzy diagnostics system for breakout prevention in continuous casting of steel

An important step in the steel production process is the casting of liquid steel into solid slabs of rectangular cross-section. This is performed in near-vertical copper molds with a coolant flowing past the entire mold outer surface, facilitating reduction in steel temperatures by more than five hundred degrees Celsius in the near-meter long mold. The heat loss is greatest in the liquid layer adjacent to mold walls, which solidifies into a thin shell of gradually increasing thickness and strength as it moves down. The formative shell may rupture due to various reasons leading to spill-out of liquid steel from mold bottom - a very undesirable phenomenon causing shutdown and production losses. This phenomenon is termed \´breakout\´. To prevent breakouts all continuous-casting systems are loaded with a \´breakout detection system\´ (BDS), which is able to evaluate real-time temperature signals emanating from mold walls and predict a potential breakout before its occurrence - bringing casting speed quickly down to near-zero to effect the prevention. This paper describes the development of a fuzzy system for breakout prevention. Although a breakout is a binary "either/or" phenomena, the complex interactions of conditions leading to/away-from it are continually evolving, and hence are implicitly conditioned to analysis using fuzzy systems. In this work, the spatial and temporal distributions of time-continuous temperatures are transformed through fuzzy systems into a time-continuous "breakoutability", i.e. a mapping of breakout tendency into a linear scale, and into a trajectory of the maximum-breakoutability point through the mold. This effectively transforms a failure prevention system into a real-time diagnostics system which also prevents failures.