Machine Health Monitoring—An Analytical Approach for Enabling Prescriptive Maintenance

Abstract

Industry 4.0 has paved the way for adopting real‐time performance monitoring of physical manufacturing systems to suggest maintenance activities. This study proposes a novel generic three‐step methodology to establish baseline performance and detect anomalies by identifying parameter deviations to enable prescriptive maintenance. Accordingly, in the first step, past machine data is utilized to generate sampling distributions for the parameters of interest. The mean of the sampling distribution establishes the ideal performance for a given parameter. In the second step, a hypothesis test is performed to detect deviation of the real‐time data from the established ideal performance. If a deviation is identified, in the third step, the exact location of an anomaly is determined. Based on the determined anomaly, appropriate maintenance activities are prescribed. The above three‐step methodology is empirically demonstrated for the health check operation of the Fluid Filling Machine, which is used in the assembly of automobiles. Data for demonstration are generated through a proposed experimental design. The demonstration establishes the workability, applicability, and managerial benefits of the proposed methodology. It is also empirically demonstrated that the proposed approach outperforms conventional preventive maintenance strategies.