25Cr2Ni4MoV锻轴制造工艺研究

对25Cr2Ni4MoV钢锻轴的冶炼、锻造、热处理工艺进行研究,并通过超声检测和力学性能检验,充分证明了其冶炼、锻造、热处理工艺的可行性。在生产实践中,成功地制造出了4支该材质的60 MW空冷发电机转子。     

碳化反沉淀法去除硫酸锰浸出液中钙、镁的研究

采用碳化沉锰法去除硫酸锰溶液中的钙、镁离子。以CO₂为碳化剂, 将硫酸锰溶液中的Mn²⁺以碳酸锰沉淀的形式从原溶液中分离出来, 然后用硫酸将沉淀物溶解, 从而达到除杂的目的。考查了CO₂流量、反应温度、pH值及反应时间对钙、镁离子去除效果的影响, 结果表明, 最适宜条件为反应温度45 ℃、溶液pH值7.0、CO₂流量2.7 L/min、反应时间60 min, 此时碳化产物中钙、镁离子含量分别为0.03%和0.01%, 达到了HG/T 2836-2011高纯碳酸锰Ⅰ型品的标准。     

立井井筒安全智能巡检机器人系统的研究

煤矿立井井筒作为矿井通往地面的主要进出口,其周围环境的安全状况以及井筒自身施工质量必须予以足够的重视。为了实时监测记录井筒环境参数、及时对井壁裂纹等状态进行评估和预警,设计了一套立井井筒安全智能巡检机器人系统。以STM32F103ZD为主控制核心,并将提升罐笼作为移动平台,通过高清摄像头和传感器等采集信息,经过WiFi无线传输至上位机,方便工作人员实时查看井筒环境状况以及相关设备运行情况,提高建设期及服役期井筒的安全保障能力。     

地面井预抽瓦斯应力-渗流耦合数值模拟研究

基于弹性力学、渗流力学等理论,建立了地面井预抽瓦斯应力-渗流耦合模型,在此基础上结合工程实例,分析了地应力对瓦斯抽采效果的影响。计算结果表明:在地面井抽采作用下,煤层瓦斯压力不断减小,且地应力越大,瓦斯压力下降速度越慢;随着抽采的持续进行,造成煤体的有效应力增加和渗透率降低,同时由于瓦斯解吸,煤层孔裂隙重新变大和渗透率增加,2种效应共同作用下煤层渗透率总体呈现非线性增加趋势;地应力对地面井抽采效率影响显著,两者呈现负相关关系,即随着地应力的增加,煤层中的基质孔隙率下降和裂隙趋于闭合,造成煤层渗透性下降,最终导致了瓦斯抽采量的下降。     

扩孔钻头在水平定向井中的应用

传统扩孔钻头在长距离水平定向井应用过程中出现泥包问题时,由于钻速变慢易造成钻进轨迹偏离原井眼,而且起钻时由于抽吸压力对井壁的作用易造成井壁压差坍塌,影响井下钻具仪器安全,增加辅助时间和钻井成本。为解决该问题,通过总结施工经验,结合数值模拟理论实验优化改进扩孔钻头,并在山西某井田煤层气水平井钻井工程项目中进行工业性试验,与传统扩孔钻头相比平均钻时减少38.9%,大大减少辅助操作时间,钻进效率提高的同时减少钻井成本。     

Experimental studies on characteristics of fire whirl in a vertical shaft

An internal fire whirl can be generated readily in a tall shaft model with appropriate gap width at one corner. Experimental study was carried out to investigate the relationship between the characteristics of an IFW and the corner gap width in a 9‐m‐tall vertical shaft model. The vertical shaft had a 2.1 m by 2.1 m square section with gasoline pool fire of different diameters burning inside. The gap width was varied to investigate its impact on fire whirl characteristics, such as flame development, swirling intensity, flame height, flame temperature, and heat release rate of the gasoline pool fire. Vigorous flame swirling motions were generated when the ratio of the gap width to the shaft section perimeter was within the range 0.16 to 0.21. From the flame streamline angle, it was observed that the swirling component was much stronger than buoyancy component near the bottom of burning region. The swirling component decreased and became roughly the same as buoyancy near the middle. Finally, it diminished to being much weaker than buoyancy near the top of the fire. These observations suggest that the Froude number Fr decreased from a large number to 1, and then continued to decrease to 0.     

When Does Communication Learning Need Hierarchical Multi-Agent Deep Reinforcement Learning

Abstract

Multi-agent systems need to communicate to coordinate a shared task. We show that a recurrent neural network (RNN) can learn a communication protocol for coordination, even if the actions to coordinate are performed steps after the communication phase. We show that a separation of tasks with different temporal scale is necessary for successful learning. We contribute a hierarchical deep reinforcement learning model for multi-agent systems that separates the communication and coordination task from the action picking through a hierarchical policy. We further on show, that a separation of concerns in communication is beneficial but not necessary. As a testbed, we propose the Dungeon Lever Game and we extend the Differentiable Inter-Agent Learning (DIAL) framework. We present and compare results from different model variations on the Dungeon Lever Game.     

改进的Faster RCNN煤矿井下行人检测算法

针对煤矿井下环境恶劣、光照差、背景混杂、行人模糊、行人多尺度等问题，提出了一种改进的Faster RCNN煤矿井下行人检测方法，使用深度卷积神经网络代替传统的手工设计特征方式自动地从图片中提取特征。利用深度学习通用目标检测框架Faster RCNN，以Faster RCNN算法为基础，对候选区域网络（Region Proposals Network，RPN）结构进行了改进，提出了一种“金字塔RPN”结构，来解决井下行人存在的多尺度问题；同时算法中加入了特征融合技术，将不同卷积层输出的特征图进行融合，增强煤矿井下模糊、遮挡和小目标行人的检测性能。实验结果表明：改进的Faster RCNN可以有效解决井下行人检测问题，在井下行人数据集上获得了90%的检测准确率，并在公测数据集VOC 07上对改进算法进行了验证。     

改进的Fast-CNN模型在绝缘子特征检测中的研究

针对目前电网巡检系统中采用红外成像检测绝缘子串特征的效果受环境影响，提出联合显著区域和Fast-CNN网络（改进后的卷积神经网络）用于绝缘子特征检测研究。显著区域检测首先采用超像素描述各区域位置的整体信息；然后基于各超像素的特征协方差信息计算各超像素的显著度得到大致显著区域；再通过区域模块化和局部复杂度对比提取显著特征，同时将2种方法提取的显著特征分别输入改进后的Fast-CNN网络进行显著区域检测，同时引入动态自适应池化模型和余弦窗处理中间层，最后通过多次迭代训练得到绝缘子特征，避免CNN模型耗时的全图搜索。将本文算法在红外图像库中进行测试，本文算法的F-Measure以及平均误差MAE均优于当前流行算法。