This presentation was made at CAASE18, The Conference on Advancing Analysis & Simulation in Engineering. CAASE18 brought together the leading visionaries, developers, and practitioners of CAE-related technologies in an open forum, to share experiences, discuss relevant trends, discover common themes, and explore future issues.

Resource Abstract

One of the most promising effects of IIoT introduction is predictive maintenance (PdM). This is because if equipment failure can be predicted with high accuracy, facility availability increases, resulting in productivity improvement.

PdM has been studied over the past decades to avoid unexpected equipment failures and to optimize the plan of operation and maintenance in large-scale plants. Due to the recent IoT trend, an environment that can easily introduce the flow from sensor installation to data collection and monitoring was provided. In addition, development of machine learning algorithms such as deep learning has dramatically progressed, and the possibility of data mining utilizing the collected big data has significantly increased. As a result, attention has also been drawn in relatively small-scale manufacturing industries that have not been able to fully deal with predictive maintenance without IoT.

However, predictive maintenance cannot be realized merely by measuring the state of the machines and accumulating the data. It is necessary to know the degradation characteristics of the mechanical elements, to grasp the influence of the operation condition, and to comprehend the correlation between the measurement data and the failure, and generally involve mid to long-term research tasks. Therefore, companies that do not have sufficient technical capabilities sometimes failed to achieve the expected effect of introducing IIoT. In traditional machine design, engineers didn't consider predictive maintenance in design phase, and predictive maintenance targeted at installed machines was thought to be an passive activity. This fact is one of the causes of making predictive maintenance difficult to realize.

In this workshop, we would like to propose a front loading design for predictive maintenance that makes maximum use of Digital Twin. Design optimization is carried out from the concept design phase considering measurement conditions, training of AI is performed by utilizing measured data in the cyber system to predict equipment abnormalities and mechanical failures, in addition, by performing the correlation analysis with the test results in the physical system, it is possible to realize an efficient optimum design aiming for predictive maintenance.