The estimation of interaction parameters in liquid iron is strongly demanded due to the difficulty of their measurements and its time consuming for enormous combinations of target solute elements in liquid iron. Therefore, several estimation models have been developed so far. In this study, the interaction parameters between metal elements and/or metalloid elements in liquid Fe are estimated by neural network computation in order to improve the estimation accuracy. The input parameters used in the neural network computation are assessed by lateral inhibition learning. The estimation results by nerural network computation with the assessed parameters reasonably agree with the recommended values in the literature.

由于测量铁液中的相互作用参数非常困难，并且需要大量的时间来溶解铁液中目标溶质元素，因此强烈需要估算铁液中的相互作用参数。因此，到目前为止已经开发了几种估计模型。在这项研究中，通过神经网络计算来估计液态铁中金属元素和/或准金属元素之间的相互作用参数，以提高估计精度。通过横向抑制学习评估神经网络计算中使用的输入参数。通过神经网络计算得出的估计结果与估计的参数合理地与文献中的推荐值相符。