Condition monitoring - fault detection and diagnosis - Dissertation Example

The use of condition monitoring can be seen as a development from preventive maintenance, which itself developed from break down maintenance. Modern process requirements demand greater availability and reliability of machines which can only be provided through accurate monitoring of machine health. This allows maintenance personnel to determine the best possible course of action based on knowledge available from condition monitoring (Mahamad, 2010). Condition monitoring has found greater favour in maintenance circles based on savings and system simplification provided by it.

Not only does condition monitoring allow the operator to make correct and on time maintenance decisions, it also allows a reduction in maintenance costs. The improvements offered in terms of greater system availability also provide direct financial benefit to processes that cannot afford to have significant maintenance delays. Overall a sizable reduction in maintenance costs and direct fiscal benefits offered by more reliable machines has pushed condition monitoring to the forefront of maintenance globally (Fuqing, 2011).

Background Condition monitoring can be carried out in a number of different ways ranging from the manual tabulation of manually measured variables to more complex and intelligent systems that offer diagnosed causes for machine wear. Over the years, condition monitoring has evolved significantly given the need to diagnose faults in larger and more dynamic industrial systems. There has been an increase in the use of artificial intelligence and a number of mathematical techniques, such as principal component analysis (PCA), in order to isolate faults and offer diagnosis for industrial systems.

Need for Artificial Intelligence (AI) Applications in Condition Monitoring AI techniques have been applied to a number of different industrial systems including condition monitoring. It must be recognised that the application of conventional techniques such as time domain, frequency domain and envelope analysis do not always yield satisfactory fault detection. In order to drive up the reliability of the fault detection mechanisms, AI and PCA are applied. More notably, neural networks and fuzzy logic have found pervasive application in condition monitoring systems.

The application of AI for condition monitoring is required in areas where analytical knowledge is difficult to come across. The use of AI allows creation of new knowledge from existing knowledge and input data from monitored variables (Shi, 2004). The use of AI and PCA techniques is required since vibration data sets contain a lot of data which results in the creation of a large set of features. Optimal feature selection is only achievable through the application of IA and PCA. A comparison of IA and PCA application versus conventional methods such as time domain, frequency domain and envelope analysis reveals that the former results in greater efficiency and savings.

The application of conventional methods requires human resources with the right expertise as well as significant time that cost the maintenance establishment significantly. In contrast, the application of IA and PCA techniques allows for much faster and more reliable fault detection without the hassle of added costs. However, it has to be kept in mind that variables measured