High voltage electrical failures are dangerous and costly events in any type of power system. The troubleshooting and diagnostic time required to identify and locate these failures can be significant. Partial discharge is one of the early warning signs for electrical degradation. In insulation systems, partial discharge typically occurs in voids located within the dielectric, at material interfaces, or along energized electrode surfaces. Effective methods for finding this failure precursor enabling circumvention of future catastrophic events are highly valuable as successful detection can improve safety, reduce service interruptions, and result in significant financial savings. Challenges arise when these events are obstructed from a direct line of sight (which is common in compact electrical systems). Conventional electrical partial discharge measurements capable of diagnosing concealed defects based on phased resolved partial discharge (PRPD) patterns require coupling devices physically connected to the circuit. This paper presents a non-invasive, real-time, method to detect and locate partial discharge and faulty insulation with potential for automated quality control of in-factory manufactured products and in-service operational devices, in contrast to post-failure assessment. This paper will cover both Alternating Current (AC), previous research, and Direct Current (DC), new research, detection methods and results.
The target of this research is to develop a machine-learning classification system for object detection based on three-dimensional (3D) Light Detection and Ranging (LiDAR) sensing. The proposed real-time system operates a LiDAR sensor on an industrial vehicle as part of upgrading the vehicle to provide autonomous capabilities. We have developed 3D features which allow a linear Support Vector Machine (SVM), Kernel (non-linear) SVM, as well as Multiple Kernel Learning (MKL), to determine if objects in the LiDARs field of view are beacons (an object designed to delineate a no-entry zone) or other objects (e.g. people, buildings, equipment, etc.). Results from multiple data collections are analyzed and presented. Moreover, the feature effectiveness and the pros and cons of each approach are examined.
The Sensor Analysis and Intelligence Laboratory (SAIL) at Mississippi State University's (MSU's) Center for Advanced Vehicular Systems (CAVS) and the Social, Therapeutic and Robotic Systems Lab (STaRS) at MSU's Computer Science and Engineering department have designed and implemented a modular platform for automated sensor data collection and processing, named the Hydra. The Hydra is an open-source system (all artifacts and code are published to the research community), and it consists of a modular rigid mounting platform (sensors, processors, power supply and conditioning) that utilize the Picatinny rail (a standardized mounting system originally developed for firearms) as a rigid mounting system, a software platform utilizing the Robotic Operating System (ROS) for data collection, and design packages (schematics, CAD drawings, etc.). The Hydra system streamlines the assembly of a configurable multi-sensor system. This system is motivated to enable researchers to quickly select sensors, assemble them as an integrated system, and collect data (without having to recreate the Hydras hardware and software). Prototype results are presented from a recent data collection on a small robot during a SWAT-robot training.