Real-time diagnosis and alarm of down-hole incidents in the shale-gas well fracturing process

Detecting down-hole incidents in the shale-gas well fracturing process plays an important role in ensuring that the fracturing operations are carried out smoothly. This paper proposes a method to monitor down-hole incidents by extracting the qualitative trend of process variables (QTPV) using qualitative trend analysis. This is based on the consideration that QTPV is similar at different magnitudes of down-hole incidents and that deviations from the normal pattern may indicate a possible incident. Based on this, this paper presents a real-time diagnosis and alarm method of down-hole incidents using a multi-class support vector machine ( MCSVM ) model for qualitative trend classification in real-time. Compared with the traditional modelling process in which process data is directly used as the input item to develop the MCSVM classifier, the proposed method can achieve higher global accuracy, as well as lower false and missing alarm rates, even with limited incident cases. Moreover, successful real-time diagnosis and alarm of down-hole incidents (cracks forming in the strata, channelling near the wellbore area, and sand plugs) are demonstrated. The results suggest that the presented method is a reasonable starting point for monitoring down-hole incidents during the shale-gas well fracturing process. This approach can be integrated into a real-time monitoring and alarm device for field application during fracturing operations.