ABSTRACT

Predicting the instant moisture content of the low-rank coals under the different drying conditions is crucial to construct the optimal system design and operation related to drying processes. Even if the thin-layer drying models are at the center of the field studies, the disadvantage of these type models is that the prediction results are valid only for the conditions of the drying experiment. Conversely, artificial intelligence models can provide accurate prediction results under a wide range of different conditions. Nevertheless, they are not practical because of their implicit forms and require both the specific software and experts. In this study, the GMDH-type neural network is applied for the first time in developing explicit model equations for the prediction of coal moisture at any time during the drying process. 223 experimental instances are used, representing coal moisture contents obtained under the different drying conditions. The considered parameters are bed height (80–150 mm), coal sample size (20–50 mm), drying air velocity (0.4–1.1 m/s), drying air temperature (70–160 °C), and drying time (0–270 minute). The developed equation is nonlinear and provides satisfactory prediction accuracy (R² is 0.96–0.99) for different drying conditions. Additionally, its usage is quite practical due to the explicit form.

摘要

预测不同干燥条件下低阶煤的瞬时水分含量对于构建与干燥过程相关的最佳系统设计和操作至关重要。即使薄层干燥模型处于现场研究的中心，这些模型的缺点是预测结果仅在干燥实验条件下有效。相反，人工智能模型可以在广泛的不同条件下提供准确的预测结果。然而，由于它们的隐含形式，它们并不实用，并且需要特定的软件和专家。在这项研究中，GMDH 型神经网络首次应用于开发显式模型方程，用于预测干燥过程中任何时间的煤水分。使用了 223 个实验实例，代表了在不同干燥条件下获得的煤水分含量。考虑的参数是床高（80-150 mm）、煤样尺寸（20-50 mm）、干燥空气速度（0.4-1.1 m/s）、干燥空气温度（70-160°C）和干燥时间（ 0–270 分钟）。所开发的方程是非线性的，并提供了令人满意的预测精度（R²是 0.96–0.99) 用于不同的干燥条件。此外，由于其明确的形式，它的使用非常实用。