ABSTRACT

The interaction of coal and oxygen releases heat that promotes oxidation process when it constantly accumulates. If heat released is more than heat dissipated during the oxidation process, the temperature increases. When the temperature of the coal reaches its ignition point, the coal ignites and potentially result in spontaneous combustion. This study evaluated and developed reliable predictive models using an artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS) and multilinear regression (MLR) analyzes based on data obtained from the proximate and ultimate analyses for coal samples collected from the Witbank Coalfields, South Africa. Thirty-five (35) coal samples data have been used, including proximate analysis, ultimate analysis, and spontaneous combustion liability indices [crossing point temperature (XPT), Wits-Ehac Index and FCC (Feng, Chakravorty, Cochrane) Index]. This study indicated that some of the samples that were categorized to have moderate and high risks according to the XPT risk rating are not in-line with those from the Wits-Ehac and FCC tests for both the experimental and published data. The results of the models showed that ANN performed better for the XPT and Wit-Ehac predictions, the ANFIS provides a better prediction of the FCC Index, while the MLR has the highest error values.

摘要

煤和氧气的相互作用释放热量，当它不断积累时，它会促进氧化过程。如果在氧化过程中释放的热量多于散发的热量，则温度会升高。当煤的温度达到其着火点时，煤点燃并可能导致自燃。本研究使用人工神经网络 (ANN)、自适应神经模糊推理系统 (ANFIS) 和多线性回归 (MLR) 分析评估并开发了可靠的预测模型，这些分析基于从 Witbank 收集的煤炭样本的近似和最终分析获得的数据煤田，南非。已使用三十五 (35) 个煤样数据，包括近似分析、最终分析和自燃易发性指数 [交叉点温度 (XPT)、Wits-Ehac 指数和 FCC (Feng, Chakravorty, Cochrane) 指数]。该研究表明，根据 XPT 风险评级被归类为中度和高风险的一些样本与来自 Wits-Ehac 和 FCC 测试的实验数据和公布数据的样本不一致。模型结果表明，ANN 对 XPT 和 Wit-Ehac 预测的表现更好，ANFIS 对 FCC 指数的预测更好，而 MLR 的误差值最高。