| dc.description.abstract | For more than a century, the induction motor (IM) has been the powerhouse
industrial applications such as machine tools, manufacturing facilities, pumping stations,
and more recently, in electric vehicles. In addition, IMs account for approximately 40%-
45% of the annual global electricity consumption. Therefore it is a critical issue to
improve IM operation efficiency and reliability. In applications, unexpected failures of
IMs can result in extensive production loss and increased costs. The classical preventive
maintenance procedures involve periodic stoppages of IMs for inspection. If such
procedures result in no faults found in the machine, as is common in practice, the
unnecessary downtimes will increase operational costs significantly. This inefficiency
can be addressed by condition monitoring, whereby sensors relay information about the
IM in real-time, allowing for incipient IM fault diagnosis. Such a process involves three
general stages:
• Data acquisition: A process to collect data using appropriate sensors.
• Fault detection: A means to process collected data, extract representative fault
features, and determine the condition of the motor components.
• Fault classification: A means to automatically classify fault data to allow
decision-making on whether or not the motor is healthy or damaged.
However, there are challenges with the above stages that are at present, barriers to the
industrial adoption of condition monitoring, such as:
• Implementation limitations of traditional wired sensors in industrial plants.
• The restrictive memory and range capabilities of existing commercial wireless
sensors.
• Challenges related to misleading representative fault signals and means to
quantify the fault features.
• A means to adaptively classify the data without prior knowledge given to a fault
classification system.
To address these challenges, the objective of this work is to develop a smart sensor-based
IM fault diagnostic system targeted for real industrial applications. Specific projects
pertaining to this objective include the following:
Smart sensor-based wireless data acquisition systems: A smart sensor network
including current and vibration sensors, which are compact, inexpensive, lowpower, and longer-range wireless transmission.
• Fault detection: A new method to more reliably extract the representative fault
features, applicable under all IM loading conditions.
• Fault quantification: A new means to transform fault features into a monitoring
fault index.
• Fault classification: An evolving classification system developed to track and
identify groups of fault index information for automatic IM health condition
monitoring.
Results show that: (1) the wireless smart sensors are able to effectively collect data from
the induction motor, (2) the fault detection and quantification techniques are able to
efficiently extract representative fault features, and (3) the online diagnostic classifier
diagnoses the induction motor condition with an average accuracy of 99.41%. | en_US |
