Research On Automatic Sorting Method Of Gangue Coal Block Based On Deep Learning

Coal accounts for about 60% of Chinese energy consumption ratio,its status is very important.In the coal mining,coal gangue will inevitably exist in the raw coal,long-term large amount of coal mining caused by the huge number of coal gangue,Shanxi Province as a large coal-producing province exists a considerable number of gangue mountain.The traditional raw coal sorting,gangue sorting method can not avoid the inclusions of a small amount of coal blocks,which causes waste of resources and environmental pollution,so the inclusions of a large proportion of coal blocks in the gangue mountain,the coal blocks from the gangue is very necessary to sort out.In view of the above problems,through analyzing the existing sorting methods with low automation and low accuracy,and combining with the current research status of deep learning in the direction of image processing,an automatic sorting method of coal gangue briquettes based on deep learning is proposed.The details are as follows:(1)According to the demand analysis of the sorting task,the experimental equipment of the automatic sorting system is built,the queuing mechanism,the image acquisition mechanism and the pneumatic sorting mechanism are designed,and the system components are selected.(2)Acquire images of gangue and lumps with typical features,image denoising and image enhancement were carried out,and the images were expanded by rotation,flipping,cropping and color jitter,and the annotation work was completed,and the dataset of gangue and coal blocks was established,which met the requirements of deep learning algorithms.(3)The common target learning algorithms are described,the YOLOv5 s algorithm is selected considering the requirements on the operation rate as well as the recognition accuracy,the network structure of its parts is introduced in detail,and the attention mechanism and loss function are improved for the demand of binary small target recognition of coal gangue sorting;(4)In order to verify the sorting effect of the system,a physical platform was built,a computer hardware system was configured,and a virtual environment for algorithm operation was built to train the algorithm,and the experiment showed that the improved algorithm was more superior;experiments were conducted for different belt speeds and different nozzle angles,which led to the range of belt speeds and nozzle angles under the best sorting effect,and a verification test was conducted,and the experiment The experiments show that the automatic sorting system designed in this paper achieves 97.2%,which greatly improves the accuracy of coal gangue sorting.