Abstract
Detecting washouts (short, enlarged hole segments) is critical for job design computations in drilling and completion. However, the large-scale raw data from a six-arm caliper can have hundreds of thousands of measurements. This means manually determiningwellbore geometry and washouts from the raw data can be time-consuming and inconsistent. The industry needs an algorithm that can detect washouts effectively and consistently, and reduce data pointswhile preserving the fundamental characteristics of the data.
In this paper, we introduce a novel approach to detect washouts with significantly reduced data points from large-scale multi-arm caliper data. The proposed algorithm employs a two-loop data points reduction technique modified from the adaptive piecewise constant algorithm (APCA). In the first loop, the raw data is reduced to an intermediate number of segments by APCA, keeping the main peaks. A new data set is then reconstructed by replacing the values in the raw data with the calculated constants, and the data isthen processed by APCA to the required number of segments. Finally hole-size and washout are calculated with equivalent volume for each segment.
Data from borehole caliper logs have been processed by the proposed algorithm to reduce data points while keeping significant washouts. Comparisons were made by running APCA with both logged field data and simulated data. Our approach detected more washouts from fluctuated data than APCA with the specified number of segments. Applications with abnormal caliper data with unevenly distributed depth intervals also illustrate the effectiveness, androbustnessof the algorithm, while the original APCA fails to determine washouts in those cases. As a data mining method for similarity search for time series, the advantage of the proposed algorithm is that it can preserve peaks while reducing data. This new algorithm could also be applied in realtime monitoring.
