Nova abordagem para auxílio ao diagnóstico de sistemas de bombeio mecânico de petróleo através da análise de segmentos das cartas dinamométricas

Abstract

This work proposes a new approach to support the diagnosing of the Sucker Rod Pump Systems through Downhole Dynamometer Cards, by analyzing segments between the points of opening and closing valves identified in Cards. Although the idea for automatic fault diagnosis in Sucker Rod Oil Pumping Systems using Dynamometer Cards has started in 1936, there is no standard computational solution yet. The Dynamometer Cards are generally analyzed by their shape as a whole to recognize anomalies in the pump operation. Recent papers have used segmentation of the Dynamometer Cards for analyzing with promising results, but treating the segments together and not singly. In general, these studies did not provide any quantitative information of severity of the anomalies diagnosed, don’t recognize concomitant anomalies and are restricted to Dynamometer Cards from a few pumping conditions. This paper proposes an approach that can provide diagnoses of concomitant anomalies, is applicable to any Dynamometer Card and potentiates the quantification of pumping conditions severities. An analysis of the shape of the pumping condition classes represented in the Downhole Dynamometer Cards (DDC), especially regarding the behavior of segments between the valve opening and closing points is presented. A survey of shape analysis in pattern recognition field is developed and a new method of identifying the valve opening and closing points is presented for the segmentation of the DDC through polygonal approximation. Then the paper presents classification tests from segments for the classes of DDC using Fuzzy logic fed with curvature statistics from the segments between the valve opening and closing points. The results demonstrate the great potential of the proposed method, providing qualitative results, allowing quantitative analyses and the recognition of concomitant pumping conditions.