The paper observes the stages of fuzzy system development, designed for forecasting the defects of steel production, including analysis of the technological processes, forecasting problem statement, construction of fuzzy models and training algorithms. The results of experimental testing of the system on one of the metallurgical facilities are considered.