基于改进DDNet的皮带输送机位移故障诊断研究

Research on Fault Diagnosis of Coal Mine Belt Conveyor Based on Improved DDNet

摘要:

针对煤矿带式输送机皮带位移故障诊断中存在局限性大、耗时长的问题,研究将故障数据进行多源异构处理,并在数据处理的基础上将边缘检测算法与深度细节网络,构建了一种结合边缘检测算法与改进深度细节网络的多源异构数据故障诊断模型。研究首先利用边缘检测算法提取输送机图像中的边缘特征,然后结合多源异构数据,并通过改进后的深度细节网络进行故障识别,并构建故障诊断模型。结果表明检测模型在皮带边缘图像数据处理的检测准确率平均值为95.27%,比目标检测算法和K最邻近分类算法的准确率高出了5.34%和10.21%。同时检测模型的图像数据查全率平均值为93.46%,比目标检测算法和K最邻近分类算法的查全率高出了4.09%和7.18%。这说明研究构建的多源异构数据故障诊断模型能够显著提升皮带位移检测的可靠性和鲁棒性,具有重要的研究价值和实际应用前景。

Abstract:

In response to the limitations and long time consumption in detecting displacement faults of coal mine belt conveyors, this study investigates the multi-source heterogeneous processing of fault data. Based on the data processing, an edge detection algorithm and deep detail network are combined to construct a multi-source heterogeneous data fault detection model that combines edge detection algorithm and improved deep detail network. The study first utilizes edge detection algorithms to extract edge features from conveyor images, then combines multi-source heterogeneous data, and uses an improved deep detail network for fault recognition, and constructs a fault detection model. The results show that the average detection accuracy of the detection model in the processing of belt edge image data is 95.27%, which is 5.34% and 10.21% higher than the accuracy of the object detection algorithm and K-nearest neighbor classification algorithm. The average image data recall rate of the simultaneous detection model is 93.46%, which is 4.09% and 7.18% higher than the recall rates of the object detection algorithm and K-nearest neighbor classification algorithm. This indicates that the multi-source heterogeneous data fault detection model constructed in the study can significantly improve the reliability and robustness of belt displacement detection, and has important research value and practical application prospects.