Stegemann J A, Buenfeld N R
Concrete Durability Group, Department of Civil and Environmental Engineering, Imperial College of Science, Technology and Medicine, Imperial College Road, London SW7 2BU, UK.
J Hazard Mater. 2002 Mar 1;90(2):169-88. doi: 10.1016/s0304-3894(01)00338-7.
Neural network models, to predict the leachate pH for single batch extraction leaching tests conducted on Portland cement pastes containing pure compounds, were constructed using existing data from the literature. The models were able to represent the known non-linear dependency of pH on acid addition, and were used to show that Cu increases, and Zn and NO3- decrease, the leachate pH for addition of 8 meq acid/g dry cement (to achieve a mid-alkaline pH). Ba, Cd, Cr(III), Ni, Pb, Cl- and OH- had no detectable effect on the acid neutralisation capacity (ANC) of the cement pastes in the concentration ranges investigated. The laboratory where testing was conducted was found to be an important predictive variable, which acted as a surrogate variable for laboratory specific variables that were not adequately reported in the literature, such as cement characteristics, sample preparation details, and leaching test and pH measurement details. This work has shown that development of good empirical predictive models for solidified product leachate pH is feasible, and is limited only by the availability of data.
利用文献中的现有数据构建了神经网络模型，用于预测对含有纯化合物的波特兰水泥浆体进行单批次萃取浸出试验时的渗滤液pH值。这些模型能够体现出pH值对酸添加量已知的非线性依赖性，并用于表明，对于添加8 meq酸/克干水泥（以达到中碱性pH值）的情况，铜会使渗滤液pH值升高，而锌和硝酸根会使其降低。在所研究的浓度范围内，钡、镉、铬（III）、镍、铅、氯离子和氢氧根对水泥浆体的酸中和能力（ANC）没有可检测到的影响。研究发现，进行测试的实验室是一个重要的预测变量，它作为实验室特定变量的替代变量，而这些变量在文献中没有得到充分报道，如水泥特性、样品制备细节以及浸出试验和pH测量细节。这项工作表明，为固化产物渗滤液pH值开发良好的经验预测模型是可行的，并且仅受数据可用性的限制。
