Electric induction motors are one of the most important and widely used classes of machines in modern industry. Large motors, which are commonly process-critical, will usually have built-in condition-monitoring systems to facilitate preventive maintenance and fault detection. Such capabilities are usually not cost-effective for small (under ten horsepower) motors, as they are inexpensive to replace. However, large industrial sites may use hundreds of these small motors, often to drive cooling fans or lubrication pumps for larger machines. Multiple small motors may further be assigned to a single electrical circuit, meaning a failure in one could damage other motors on that circuit. There is thus a need for condition monitoring of aggregations of small motors. We report on an ongoing project to develop a machine-learning-based solution for fault detection in multiple small electric motors. Shallow and deep learning approaches to this problem are investigated and compared, with a hybrid deep/shallow system ultimately being the most effective.
1 November 2023 in IEEE Transactions on Power Delivery
Underground power cables are a widely-deployed class of assets in electric power distribution systems. They are robust in adverse weather conditions and are an aesthetically pleasing alternative to overhead power lines. Unfortunately, the physical observability of underground cables is poor. Current methods for detecting and localizing defects known to progress to failure are cumbersome and expensive, both financially and operationally. In-service cable failures can lead to lengthy power outages and financial losses and can endanger utility staff and the general public. There is thus a need for non-invasive condition monitoring techniques that will allow for early detection of defects likely to progress to a fault, so cost-effective repair or replacement interventions can be planned. In this study, insulation overheating is specifically considered. Thermal damage of XLPE insulation is a prominent damage mechanism for extruded dielectric cable and is thus well studied and understood. A defect detection solution combining powerline modems with semi-supervised anomaly detection for underground cables is proposed. The performance of five well-regarded machine-learning-based anomaly detection algorithms is analyzed through simulation of a real-world 15 kV distribution network. The best-performing algorithm studied had a false alarm rate of under 0.1% with no missed alarms across a range of damage scenarios.
Anomaly detection in data streams (and particularly time series) is today a vitally important task. Machine learning algorithms are a common design for achieving this goal. In particular, deep learning has, in the last decade, proven to be substantially more accurate than shallow learning in a wide variety of machine learning problems, and deep anomaly detection is very effective for point anomalies. However, deep semi-supervised contextual anomaly detection (in which anomalies within a time series are rare and none at all occur in the algorithm’s training data) is a more difficult problem. Hybrid anomaly detectors (a “normal model” followed by a comparator) are one approach to these problems, but the separate loss functions for the two components can lead to inferior performance. We investigate a novel synthetic-example oversampling technique to harmonize the two components of a hybrid system, thus improving the anomaly detector’s performance. We evaluate our algorithm on two distinct problems: identifying pipeline leaks and patient-ventilator asynchrony.