Condition monitoring, fault diagnosis, fault classification or fiding Dissertation

Condition monitoring, fault diagnosis, fault classification or fiding fault for machenary - Dissertation Example In recent years, there has been a growing trend to introduce more intelligent methods in order to deal with condition monitoring and fault classification for machines (Mills, 2010). The realm of artificial intelligence and its application may be infant as yet but still involves the application of various methods and techniques for achieving desired ends. The current research will look into various artificial intelligence methods that have been applied to the condition monitoring and fault diagnosis for a reciprocating air compressor based on emerging and already developed methods and techniques. 1.2 Artificial IntelligenCe Based Methods It is possible to solicit problems in plant machinery using vibration signals that can be processed to reveal a multitude of information relating to the machine and its components as well as their operation (Wang & Chen, 2011). Given that condition monitoring and diagnosis relies largely on vibration feature analysis, it is important to extract the vi bration signals at every state change that the machine experiences (Lin & Qu, 2000) (Wang & Chen, 2007). Extracting vibration features can often be difficult since the measured vibration patterns tend to contain a large amount of noise that must be filtered out (Wang & Chen, 2011). ... The application of these techniques would allow for both pattern recognition as well as automated fault diagnosis. A number of different researches have been carried out in order to deal with condition monitoring and fault diagnosis of plant machinery that relies on discriminating fault types from a common pool of fault types based on the available vibration data. Theoretically, such an approach may make a lot of sense but practical application of such techniques is hindered by ambiguous diagnosis problems. It is possible that first layer symptoms may be similar for a number of different faults and it is also possible that first layer symptoms may have similar values in different states. The situation is complicated by the fact that there are no definite relationships between symptoms and fault types for plant machinery. The added complexity of plant machinery and the various interacting components means that the overall fault states are enormous to say the least. It is not possible to rely on one or on a number of different symptom parameters that could be utilised to track down faults, supposing that each fault occurs independent of others. This situation is complicated all the more when faults tend to occur simultaneously and the application of theoretical frameworks tends to fail altogether or in large part (Mitoma et al., 2008) (Wang & Chen, 2008). A number of different methods and techniques have been applied in recent researches in order to solicit vibration feature extraction and analysis for accurate and reliable condition monitoring and fault diagnosis. These techniques and methods could be classified as (Wang & Chen, 2011): wavelet transform; rough sets; neural networks; sequential fuzzy inference;