The purposes are to effectively solve the graphic design problems, develop an easy-to-use supporting design program, and make graphic design more reliable and accurate. Based on the analysis of the current graphic design framework, graphic design data are obtained from the network social media. A system of surrounding rock classification and support optimization design is developed by a deep neural network structure. The model’s effectiveness is verified by more than 3000 road conditions data. The results show that the three-layer network’s errors are 0.0062 with a training time of 12,455, and the five-layer network’s errors are 0.00019 with a training time of 69,895. With the input layer, hidden layer, and the output layer of 8, 15, and 5 respectively, the model performs best. In the deep learning algorithm, the deep backpropagation neural network (Deep BPNN) can obtain the best training effects with less training time. Therefore, the roadway drawing system’s application based on the deep learning algorithm to the roadway support design can improve design efficiency and scientificity.

目的是有效解决图形设计问题，开发易于使用的辅助设计程序，并使图形设计更可靠，更准确。在对当前图形设计框架进行分析的基础上，从网络社交媒体获取图形设计数据。通过深度神经网络结构开发了围岩分类与支护优化设计系统。超过3000个路况数据验证了该模型的有效性。结果表明，三层网络的误差为0.0062，训练时间为12455，五层网络的误差为0.00019，训练时间为69895。在输入层，隐藏层和输出层分别为8、15和5的情况下，模型表现最佳。在深度学习算法中，深度反向传播神经网络（Deep BPNN）可以以更少的训练时间获得最佳的训练效果。因此，基于深度学习算法的巷道制图系统在巷道支护设计中的应用可以提高设计效率和科学性。