Application of selected method of anomaly detection in signals acquired during welding process monitoring

Condition monitoring of machinery and processes plays an important role in technical diagnostics. Detection of any changes in diagnostic signals that deviate from values understood as 'normal' could be treated as early detection of symptoms of defect formation or first symptoms of failure. Early detection of such events occurring in signals recorded during processes or machinery operation monitoring could prevent the machine from being damaged or could reduce the risk of downtime during production process. Different kinds of anomalies could be observed in signals acquired. Examples are point, contextual or collective anomalies. Depending on the type of the anomaly, various algorithms can be used. It is important to underline that there is no universal method for detection of all anomaly types. Selection of a proper method strongly depends on the nature of the process or phenomenon being observed. In this paper anomaly detection in signals acquired during monitoring of welding process is studied and sample applications of the selected statistical method are given. In the research undertaken two groups of signals were considered. In the first group were the welding process parameters analysed in time domain (voltage and current). In the second group were the results of statistical image analysis of the welding arc area.