In a typical predictive maintenance framework, embedded system models play
a key role for producing (quality) forecasts, for indicating arising problems and
faults at an early stage, or for conducting any deeper diagnosis about upcoming
expected (as predicted) anomalous process behaviors in various forms. The high
dynamics in today’s processes or parts of processes often has the effect that already
modeled/learned dependencies become outdated, which requires system models to
self-adapt over time in order to maintain their predictive performance and to expand
their “knowledge” and “validation range.” This is hardly considered in the current
state of the art of predictive maintenance; therefore, it is a central aspect in this
book to show new trends in this direction—in fact, most of the chapters are dealing
with (data-driven) modeling, optimization, and control (MOC) strategies, which
possess the ability to be trainable and adaptable on the fly based on changing system
behavior and nonstationary environmental influences.
Apart from this, several new applications in the context of predictive maintenance
as well as combinations of MOC methodologies to successfully establish predictive
maintenance are demonstrated in this book. According to the essential steps in
predictive maintenance systems from early anomaly and fault detection during the
process through the prognostics of eventually arising problems in the (near) future
to their diagnosis and proper reactions on these (through optimization, control for
repair, and self-healing), the book is structured into three main parts, where in each
of them, important real-world systems and application scenarios are discussed:
• Anomaly detection and localization
• Prognostics and forecasting
• Diagnosis, optimization, and control