Predicting Compressor Valve Failures from Multi-Sensor Data

Abstract

Abstract As the number of sensors deployed in the oilfield increases, there is a corresponding need to develop methods for fast, automated processing of large-scale sensor data streams. This problem is aggravated when the streams of sensor data are from multiple sensors with different characteristics and applicability to specific tasks of interest to the oilfield operator. Predictive analytics and data mining approaches have been proven to improve production efficiency, reduce downtime and identify safety hazards in real-world scenarios. One of the most common failures in rotating equipment such as compressors is the breakdown of valves. This issue is of great value because a large proportion of production is dependent on rotating equipment. The aim of our undertaking is to find signature(s) in sensor data collected from compressors, which are predictive of valve failures. The first step towards achieving this is to rank the sensors themselves and use data from only those sensors that provide useful information. The resulting information can be used to prioritize and monitor maintenance schedules for compressors, which are often on remote platforms. The data used in our evaluation is from a large number of sensors that measure various physical properties of compressors, ranging from compressor vibrations and motor winding temperatures to pressure and temperature for both suction and discharge at the various compression stages. We frame this failure prediction problem as a feature selection and time-series classification task. We use several evaluation methods for feature selection to identify the highly ranked sensors, and obtain insights about them. Once feature selection is complete, we propose to build a classifier based on time series shapelets. In particular, we adopt feature selection methods to automatically rank sensor streams in order of usefulness for the specific prediction tasks. Our results demonstrate that feature selection and time series approaches can be extended to handle multi-sensor Big Data streams from the oil and gas industry provided that accurate labeling of these datasets is possible.