Automatic Fault Detection for Resistivity Systems in Logging-While-Drilling Tools

Today’s oil and gas industry frequently uses logging-while-drilling (LWD) tools to operate in extreme environmental conditions. These conditions include elevated temperature, vibration, and pressure, which can result in accelerated tool degradation rates and possible failures. Detecting such failures before conducting a new drilling job is of great significance to guarantee the efficiency and success of the job. Previously the fault detection task required field engineers to check relevant signal channels manually, which is inefficient, labor-intensive, and could even be inaccurate sometimes. Considering this challenge, the authors developed a data-driven model that enables automatic fault detection of the resistivity system, a critical subsystem of LWD tools. Specifically, this model is based on the support vector machine method using features extracted from a few signal channels selected by a domain expert. Experimental results on actual field data show that the model is effective.