Original Article

An expert suggestıon system desıgn to optımıze oxygen enrıchment performance of ıronmakıng process ın combınatıon wıth hot blast flow predıctıon usıng neural and fuzzy-drıven algorıthms

¹ Honeywell International ME, Emaar Business Park 2/3, Sheikh Zayed Road, P.O. Box 232362, Dubai, UAE
² Istanbul Gedik University, Cumhuriyet Mahallesi, Ilkbahar Sokak, No: 1-3-5 Yakacik, 34876, Kartal, Istanbul, Turkey

^* e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 13 November 2025
Accepted: 8 May 2026

Abstract

Blast furnace (BF) process ıin large-sized integrated complex plants is dependent on various process aspects due to sensitive heat balance and thermodynamics requirements. Operational parameters such as hot blast pressure, permeability, hot blast temperature, top gas pressure and temperature, wall pressures and temperatures, ore-to-coke ratio, flame temperature, additional steam and oxygen injections, etc. are aimed to keep balanced without any sudden fluctuations, movement or interruptions by experience and heuristic approach. In this paper, hot blast flow of a BF is modeled using ANN (Artificial neural networks) by selecting input parameters and time-series based statistical ARIMA (Autoregressive integrated moving average) model is applied using the same data and input-output set to compare the predictions success using performance criterion, R², RMSE and MAPE. Secondly, a fuzzy logic model is developed to support oxygen enrichment decisions and model evaluates metal temperature, combustion and system balance trends, and makes oxygen increase, decrease or hold decisions. The experiment output reveals that ANN is very accurate to track hot blast flow values and shows better performance than ARIMA and proposed fuzzy-driven expert system could identify oxygen enrichment actions as next step to have efficient and cost-effective operation. The primary scientific novelty of this work is threefold: (i) for the first time in the open literature, a tightly coupled architecture is presented in which a data-driven ANN predictor for hot blast flow feeds directly into a real-time fuzzy expert system for oxygen enrichment control, creating a closed-loop advisory pipeline; (ii) the hybrid system is benchmarked against a statistical ARIMA baseline using three independent performance metrics grounding the evaluation in real plant conditions rather than simulated data; and (iii) hierarchical Mamdani type fuzzy inference architecture with metallurgically subsystems. Decision distributions yield 54.1% “Hold”, 37.8% “Decrease”, and 8.1% “Increase”, demonstrating that the model adopts safe and stable control approach.