Publication Details
Abstract
Predictive maintenance has gained importance in modern times as a crucial approach to maintaining mechanical systems in order to improve reliability and reduce maintenance costs. Conventional approaches such as corrective and preventive maintenance might fail to mitigate potential problems caused by unpredictable failures and progressive degradation of system components. In recent times, there have been significant advances made in predictive maintenance using statistical and machine learning approaches. Statistical approaches provide useful techniques that help analyze reliability and predict failures, while machine learning algorithms can help identify useful patterns in large sets of data.
The following review article provides a detailed discussion on the statistical and machine learning approaches used for predictive maintenance of mechanical systems. This is followed by a discussion on the evolution of maintenance strategies and the adoption of predictive maintenance approaches. Finally, some important statistical approaches relevant to predictive maintenance are discussed here. The review also covers some of the important machine learning algorithms for predictive maintenance, including support vector machines, random forests, artificial neural networks, and deep learning techniques.
The review article discusses the various types of sensor data used in the prediction maintenance of mechanical systems, including vibration, acoustic, temperature, and operational load data. It also includes a comparative study of statistical and machine learning techniques, along with the advantages and disadvantages, and the application of these techniques in various mechanical systems. At the same time, the review identifies the problems in the prediction maintenance methodology and the future direction in the field.