In many fields, different elements (mechanical, electronic, electromechanical, etc.) come into play independently to ensure the overall operation of devices/machines: electrical cables, gears, bearings, pulleys, etc. In maintenance operations, the signals from these different elements in operation are often a mixture of multiple contributions, the level of complexity of which may vary, for example, depending on the measurement point, during the data acquisition stage. There may be a number of reasons for wanting to access only the signal from a particular component (e.g., to monitor service life, diagnose the faulty part and/or predict the time remaining before a serious breakdown), rather than all the contributions measured. It is therefore essential to have a range of tools to enable us to remove harmful signals (sometimes called noise). This article presents some techniques for separating discrete and random components.