An Overview on the Use of Machine Learning Algorithms for Identifying Anomalies in Industrial Valves

Valves used in industrial systems may present anomalous behaviour or eventually fail throughout their useful life. The automatic detection of these anomalies and failures is quite important nowadays, specially due to the increasing movement of industry 4.0 and Digital Twins technologies. In order to detect their anomalies or failures, data captured by sensors in these valves can be used for analysis and prediction tasks. More specifically, these time series data are analyzed and used for constructing machine-learning based models for predicting, which may contribute to better preventive maintenance and longer valve life. The objective of this work is to present the results obtained when executing a Systematic Literature Review (SLR) in order to identify the state of the art on using Machine Learning (ML) algorithms and methods to identify anomalies in industrial valves. Case studies could be identified on different types of valves, including hydraulic, solenoid, gas, nuclear and compression valves. This indicates a long future to develop and propose solutions in this field of science. An increasing focus on the use of deep learning models could also be observed, although there are also simpler methods under scrutiny. Finally, it could be observed that it is quite difficult to find available datasets in order to replicate the obtained results, as well as advance in new ideas and methods in this area.