Tagarchief: TUE-model

AI and condition monitoring

In recent history, many attempts have been made to apply Artificial Intelligence (AI) in maintenance. Essentially the use of AI is an attempt to replace human intelligence with machine intelligence. The objective is to achieve better maintenance. There has been vast interest in the application of AI in the maintenance area as witnessed by the large number of publications. The continuous research in this area imply that while theoretical situations solutions are available, practical and real life solutions still not be found. This may be attributed to the fact that proposed solutions fit for well-defined problems and that these are based accurate data.

There are two approaches for Preventive Maintenance planning: the engineering approach (Gits 1984 – TUE Model) and the mathematical approach. The last has an emphasis on developing optimal maintenance intervals. Condition monitoring adds, as a continuous process, a new dimension to these approaches. Condition monitoring requires rules based on a model (based on a known failure pattern) for planning and monitoring. Since the resurge of AI in de mid-1980 researchers have considered the application of AI in this field.

With the development in IT the last decades, many organizations have systems (ERP, CMMS) to collect and retrieve data on maintenance actions. Although the stored historical data is potentially useful to improve useful to improve maintenance actions, nowadays the data is mainly used to produce management reporting. The following difficulties hinder new developments to learn from this data.

  • The vast number of components, (sub-)systems and therefore the wide variety of maintenance policies and situations.
  • The lack of familiarity to modelling in addition to maintenance expertise
  • The physical access from the analytical software with the maintenance environment
  • The dynamics of the systems itself (other components, other operational conditions or change in requirements or purpose

Other issues are the quality of the existing data, the (human) ability to recognize patterns, the learning capacities (ML- Machine Learning), the presentation of findings in a visual attractive format.

In addition to the problems above another problem ads complexity to the role of AI in optimizing maintenance tasks. At a higher level, management is concerned with the effectiveness of systems, modes and failures and in return the type of maintenance actions to improve operations. At the lower decision level, maintenance engineers are concerned with efficiency issues, e.g. intervals.