废弃橡胶-砂混合物热传导特性与预测模型研究

Thermal Conduction Behavior and Prediction Model of Scrap Rubber-Sand Mixtures

为研究添加废弃橡胶对砂导热性能的影响规律,通过室内热探针试验测试了不同制备状态下废弃橡胶-砂混合物的导热系数,研究了导热系数随橡胶掺量、含水率和干密度变化的变化规律,分析了多种因素对橡胶-砂混合物热传导特性的综合影响,并在所得试验数据基础上,建立了基于人工神经网络技术的导热系数预测模型,对模型的有效性和适用性进行了验证。结果表明:添加橡胶颗粒可显著降低砂的导热性能,橡胶-砂混合物的导热系数随橡胶掺量的增加而降低,降幅与含水率密切相关;增加含水率会提高橡胶-砂混合物的导热系数,超过临界含水率后,导热系数基本不变,本文橡胶-砂混合物的临界含水率约为8%;增大干密度有利于砂、橡胶颗粒之间的良好接触,提高热量在混合物中的传输效率,进而增大橡胶-砂混合物的导热系数;人工神经网络计算模型可有效预测橡胶-砂混合物的导热系数,预测值与实测值的相关系数R~2大于0.85,相对误差绝对值小于10%,模型预测精度满足工程设计的需求且具有良好的适应性。建议今后开展橡胶、砂颗粒形貌对其混合物导热性能的影响研究,以助于进一步揭示热量在多孔颗粒介质中的传递机理。

In order to clarify the influence law of adding scrap rubber on the thermal conduction of sand,this paper conducts a series of thermal probe tests on the rubber-sand mixture samples with varied preparation conditions.The paper investigates the effects of rubber content,moisture content,and dry density on the measured thermal conductivity,and analyses the comprehensive effects of some influence factors on the thermal conduction behavior of rubber-sand mixtures.Based on the obtained data set,the paper develops a prediction model within the framework of artificial neural network for thermal conductivity of rubber-sand mixtures,and then demonstrates its validity and applicability.The results are as follows:an addition of shredded rubber can obviously reduce the thermal conduction ability of sand,and thermal conductivity of rubber-sand mixtures decreases with an increase in rubber content and the magnitude of this reduction closely depends on the moisture content;thermal conductivity of rubber-sand mixtures increases with the increasing of moisture content,and once the moisture content exceeds the critical value,the thermal conductivity keeps almost a constant.The critical moisture content of the investigated rubber-sand mixtures in this study is approximately 8%.Increasing dry density is beneficial to obtain good contact between particles,resulting in an improvement of heat transfer efficiency in the mixtures and an increase in thermal conductivity of the mixtures.The proposed ANN prediction model performs well in predicting thermal conductivity of rubber-sand mixtures with correlation coefficient R~2 higher than 85%and the absolute value of relative error lower than 10%.This prediction accuracy meets the requirements of engineering design and exhibits a good adaptability.We suggest that the effects of particle morphology on thermal conductivity of rubber-sand mixtures is warrant to be conducted in future,which is helpful to further explore the heat transfer mechanism of porous particle matrix.