Abstract: The invention relates to using autoencoder-derived features for predicting well failures (e.g., rod pump failures) using a machine learning classifier (e.g., a Support Vector Machine (SVMs)). Features derived from dynamometer card shapes are used as inputs to the machine learning classifier algorithm. Hand-crafted features can lose important information whereas autoencoder-derived abstract features are designed to minimize information loss. Autoencoders are a type of neural network with layers organized in an hourglass shape of contraction and subsequent expansion; such a network eventually learns how to compactly represent a data set as a set of new abstract features with minimal information loss. When applied to card shape data, it can be demonstrated that these automatically derived abstract features capture high-level card shape characteristics that are orthogonal to the hand-crafted features.

Abstract: Methods and systems for predicting failures in an artificial lift system are disclosed. One method includes extracting one or more features from a dataset including time sampled performance of a plurality of artificial lift systems disposed across a plurality of different oil fields, the dataset including data from failed and normally operating artificial lift systems. The method also includes forming a learning model based on identified pre-failure signatures in the extracted features, the learning model configured to predict a failure of an artificial lift system based on observation of one of the identified pre-failure signatures in operational data received from the artificial lift system.

Abstract: An apparatus and method is described for multicasting an inlet data cell, received as part of a multicast request, through a self-routing multistage routing network. Two algorithms are disclosed which use at most three passes to perform any arbitrary multicast. In the first pass, data is usually routed from the source to a set of consecutive outputs. These outputs are used as inputs in the second pass for routing to the destination cubes, and in some cases, a third pass can be used to route the remaining cubes. Since these cubes constitute a partition of the original multicast set, the multicast is completed when each of the cubes has been successfully routed. The routing algorithms guarantee that those cubes used in each pass through the network do not have internal blocking. The criterion to select a multicast routing algorithm for a particular application is to keep a appropriate balance between the control simplicity, the number of passes and the number of internal links used by the algorithm.