The presented state of the art as well as the practical circumstances lead to maintenance systems where one
standard strategy (i.e. predictive maintenance) is followed. Regarding the case of offshore wind turbines for
instance, a yearly inspection with an extensive maintenance plan is used to ensure the predictive recognition of
future faults. Nevertheless, failures often lead to downtimes of the system and to troubleshooting with corrective
measures. Moreover, the monitoring of the technical condition of systems or components by sensor technology
generally just enables the assessment of a subset of data that would be needed for a holistic perspective.
Regarding the methodological basics and the main concept of the so-called “preacting” maintenance approach,
the fundamental level is still based on condition monitoring and data processing. Besides the technical perspective,
economic data (e.g. historical data of the maintenance measures) or environmental data (like weather in the case of
offshore wind turbines) have to be considered. Condition monitoring and data processing will lead to a deep
understanding of the current condition of technical systems. Hereby, they achieve the general ability to inquire
maintenance tasks related to their particular requirements.
Based on that, the preactive paradigm includes three further objectives of traditional maintenance strategies.
Corrective maintenance tasks or troubleshooting are supported by recognition of failures at an early stage so that
planning and controlling as well as scheduling of tasks and according resources are enabled. Moreover, preventive
maintenance is followed for some components essentially with a linear wear-out curve. Those tasks should be
assessed under cost-/risk considerations. Economic effects take place in case those preventive tasks are scheduled
together with other ones (e.g. troubleshooting). Finally, the maintenance and inspection based on fixed intervals can
be continued, which is not typical for general condition-based approaches. Data and information derived from the
previous described basis should enable dynamic maintenance and inspection scopes according to the particular
condition or requirements of considered technical systems. From the methodical perspective, data and information
have to be processed in order to schedule operative executions of maintenance measures according to the following
assumptions:
x Maintenance measures are put into an optimal order according to the priority, which has been built under
consideration of different criteria, e.g. earliness-tardiness costs and risks, available resources, accessibility, etc.
Considered systems gain the ability to know their own state by themselves and negotiate with resources regarding
their local and opportunistic objectives.
x The scope of the maintenance measures will be built dynamically and ad hoc under consideration of aggregating
different measures to the same system or arranging the tasks of inspection measures for instance based on the
measured and assumed conditions.
336 Marco Lewandowski and Stephan Oelker / Procedia Technology 15 ( 2014 ) 333 – 340
Finally, heterachical planning and control leads to a positive emergence of the overall social-economic
maintenance system. In the following, concrete methodical approaches are presented for the key elements that are
necessary to build such a system, namely the condition-monitoring system, the assessment of condition and priority
and the agent-based negotiation ability of involved actors.
3.1. Condition-monitoring systems for data acquisition
Nowadays, condition-monitoring systems collect much data from different sources. From the maintenance point
of view, the data in general will be used to determine the condition of technical systems. Most often, limits are defined
in order to identify problematic situations or failures out of the data stream. Regarding the planning and controlling
of maintenance measures at an early stage, data and information are usually too late so that, in reality,
measures are rather corrective than preventive.
Condition monitoring in detail includes the diagnosis, which retrospectively recognizes patterns in terms of
symptoms of failures, classifies the detected errors and analyzes the causes. However, there is a necessity of forward-looking
prediction, i.e. the estimation of further consecutive faults. This is done on the basis of the analysis of
time series and their forecasting. The most important aggregate value that results from this is residual life estimation.
The forecast has a feedback loop to the diagnostic process. While the diagnosis can operate as a single resultoriented
process, the isolated application of the prognosis usually is not possible. There have to be first outcomes of
the diagnosis that serve as an input variable for forecasting.
Targets related to the concept of preacting maintenance are the design of an integrative framework to reflect
current technologies and numeric