Abstract
Petroleum and natural gas are pivotal in the energy sector, with drilling as the primary extraction method involving machine-driven drill bits for excavating underground and retrieving resources. Similar to various industrial processes, drilling efficiency is significantly affected by vibration-related issues, influencing petroleum extraction outcomes. To systematically identify vibrations during drilling, we developed a data collection system covering both surface and downhole components. The surface component primarily records parameters observed during drilling, while the downhole part integrates into the drilling tool, positioned near the drill bit to capture three-axis vibration signals. The amalgamation of data from both surface and downhole components facilitates the annotation of harmful vibrations. Using LSTM-FCN, we enhance vibration identification by processing surface and downhole data. Our SampleFusion (SF) algorithm further improves recognition by combining LSTM for surface and FCN for downhole data. This integrated approach enhances real-time monitoring and classification of drilling vibrations, contributing to industry efficiency and safety.