Abstract In recent years, the use of solid desiccant wheels has become attractive not only for air-conditioning applications, but also for food drying processes and storage due to their capacity to use waste heat in order to meet dry and hot air demand. It is very important that solid desiccant wheels be modelled for different purposes in such a way that the equipment can be well analysed in various systems. Modelling the solid desiccant wheel is a difficult and complex process because of the coexisting heat and mass transfer. In this study, six Artificial Neural Network (ANN) models with various activation functions and Multiple Linear Regression (MLR) models with six different structures have been formed to observe the process air outlet conditions of the solid desiccant wheel, and compared with each other to identify the suitability of the use of these models. In comparison, R2, RMSE and MAE values were taken into consideration as performance criteria. At the end of the study, ANN models were observed to provide better convergence than MLR models. The best convergence for the process air outlet conditions was provided by the ANN-V model. Of all the MLR models, the best convergence was provided by MLR-VI model.

中文翻译：

---

多元线性回归 (MLR) 和人工神经网络 (ANN) 技术模拟固体干燥剂轮的比较研究

摘要 近年来，固体干燥轮的使用不仅对空调应用具有吸引力，而且由于它们能够利用废热以满足干燥和热空气需求，因此在食品干燥过程和储存中也很有吸引力。为不同目的对固体干燥剂轮进行建模非常重要，这样可以在各种系统中很好地分析设备。由于同时存在传热和传质，对固体干燥剂轮进行建模是一个困难而复杂的过程。在本研究中，形成了具有各种激活函数的六个人工神经网络（ANN）模型和具有六种不同结构的多元线性回归（MLR）模型来观察固体除湿轮的工艺空气出口情况，并相互比较以确定使用这些模型的适用性。相比之下，R2、RMSE 和 MAE 值被视为性能标准。在研究结束时，观察到 ANN 模型比 MLR 模型提供更好的收敛性。ANN-V 模型提供了工艺空气出口条件的最佳收敛。在所有 MLR 模型中，MLR-VI 模型提供了最好的收敛性。