Relationships among bioleaching performance, additional elemental sulfur, microbial population dynamics and its energy metabolism in bioleaching of chalcopyrite

To estimate the relationships among bioleaching performance, additional elemental sulfur （S0）, microbial population dynamics and its energy metabolism, bioleaching of chalcopyrite by three typical sulfur- and/or iron-oxidizing bacteria, Acidithiobacillus ferrooxidans, Leptospirillum ferriphilum and Acidithiobacillus thiooxidans with different levels of sulfur were studied in batch shake flask cultures incubated at 30 °C. Copper dissolution capability （71%） was increased with the addition of 3.193 g/L S0, compared to that （67%） without S0. However, lower copper extraction was obtained in bioleaching with excessive sulfur. Microbial population dynamics during chalcopyrite bioleaching process was monitored by using PCR-restriction fragment length polymorphism （PCR-RFLP）. Additional S0 accelerated the growth of sulfur-oxidizing bacteria, inhibited the iron-oxidizing metabolism and led to the decrease of iron-oxidizing microorganisms, finally affected iron concentration, redox potential and bioleaching performance. It is suggested that mixed iron and sulfur-oxidizing microorganisms with further optimized additional S0 concentration could improve copper recovery from chalcopyrite.

基于自适应锚框的裂缝目标检测算法研究

随着交通的发展,桥梁在运输过程中扮演着越来越重要的角色,桥梁也更加多样化。因此面对大量工况不同的桥梁,发展一种能便捷学习新工况的智能化裂缝检测技术显得尤为重要。为提高目标检测算法的准确率和效率,本文将裂缝原始图像切分成3种不同分辨率和尺寸的切片,训练网络识别不同尺寸的裂缝。同时为了增加算法的后续拓展性,设计一种根据训练集标注尺寸自适应调整锚框的手段,让算法在后续使用过程中针对不同工程情况需要增加训练数据时,能直接添加数据进行训练,自动调整最佳锚框尺寸,使该算法在实际使用中具有学习改进的空间。与原始YOLOv3网络和文献中的算法对比,本文算法的精确度平均达到91%以上且扩展性更好。

先验信息强化的混凝土裂缝深度学习级联检测算法研究

为提高混凝土桥梁复杂表面的裂缝检测精度,提出一种基于深度学习的先验信息强化的裂缝级联检测算法。算法采用级联检测的方法将2个神经网络级联来检测裂缝;先验信息强化包括标准化和边缘强化,制作裂缝标准化切片时将裂缝旋转至水平走势并将质心放置在切片中心,同时根据灰度梯度相似性去除属于一般背景的切片;将切片进行裂缝边缘先验信息强化处理,以此人为地突出特征,降低网络训练难度。分别对比不采用先验信息强化和级联检测的网络结果,评价网络检测性能。实验结果表明,对混凝土桥梁表面裂缝检测的精确率和召回率达到89.64%和98.02%,能较好地避免背景中干扰物的影响。