AI-Driven Predictive Maintenance Technology for Intelligent Manufacturing: A Review
TL;DR
The whole closed-loop technical architecture of "sensing - evaluation - determination - actuation" and the operating logic of the five basic stages of data acquisition, improvement, analysis, strategic planning and feedback are introduced here, along with the research and application results of some representative machine learning and deep learning architectures for fault diagnosis and remaining useful life estimation.
Abstract
Based on the development trends and application cases of artificial intelligence-powered predictive maintenance technology in intelligent manufacturing, a general analysis will be carried out in this paper. The old way of fault repair and time-based preventive maintenance has the following problems: there will be significant business disruption due to unexpected failure; maintenance will be carried out unnecessarily; and resources will not be used efficiently. With the help of the new IoT platform and AI algorithm, condition monitoring for the maintenance of industrial equipment has begun to be used in practice. The whole closed-loop technical architecture of "sensing - evaluation - determination - actuation" and the operating logic of the five basic stages of data acquisition, improvement, analysis, strategic planning and feedback are introduced here, along with the research and application results of some representative machine learning and deep learning architectures for fault diagnosis and remaining useful life estimation. At the same time, it is pointed out that there are still issues with the small-sample generalisation ability and model transparency of this method, and this provides a necessary direction for further improvement and practical application of industrial smart operation and maintenance platforms.