“…Typically, data-driven PdM relies extensively on historical monitoring data to observe the evolution of the equipment from a safe initial condition to the break of the equipment part/machine failure. Commonly used features for constructing the prediction model include property features (e.g., vibration signal data, temperature, current, and voltage) ,,,,,,− ,− and historic process features (e.g., recipe and log data). ,, Currently, data-driven PdM has been widely applied to a variety of equipment from various industries, such as bearings, filaments, turbines, motors, gearboxes, and compressors. ,,,,,, The most employed algorithms (one study can apply more than one ML algorithm) are RF (35.7%) − ,,,, and SVM (32.1%), ,,,,,,, while DL, such as long short-term memory (LSTM), has become increasingly popular in recent years (32.1%). ,− ,,, Such proposed models can provide a prediction accuracy range of 77.2% to 99.84% for identifying malfunctions which significantly outperforms traditional models. ,,, …”