The MNN can be used for on -line learning of the plant's dynamic characteristics and provide a kind of adaptive control strategy which can improve the whole control system's performance.

The method of establishing on-line temperature prediction model, set point model and parameter tracing self-learning model of computerized heat recovery processing control system is introduced.

The result of simulation shows that this neural network fuzzy controller features self-learning and self-adaptive capabilities, and the purpose of on-line control is accomplished.