Leveraging Data Consolidation and Event Driven Systems to Provide Predictive Diagnostics on the Factory Floor

Maintenance activities on a plant floor result in lost production and capacity, and consequently lost money. Unscheduled maintenance events are especially costly and can generate costly ripple effects in other parts of the factory up and down the supply chain. The University of Michigan - Engineering Research Center (ERC) is currently working on a project with a major automotive manufacturer, whose objective is providing a factory-wide predictive diagnostics infrastructure by deploying an open-architecture event-driven software control system. The control architecture will link equipment data collection, equipment control and equipment & tool maintenance capabilities, and will provide for coordination of resources through event driven systems to both reduce unscheduled downtime and lessen mean-time-to-repair (MTTR). The proposed solution uses data that is currently collected from the plant floor; this data, when consolidated and collectively analyzed, will be utilized to enhance the predictability of maintenance events. This approach enables the leveraging of the existing plant data collection infrastructure into the control solution architecture. The first step of the project is a historical study of the data from the different plant floor systems to identify trends leading to failures. The next project step is the implementation of an event driven control solution that utilizes the Event Condition Action (ECA) paradigm with ECA control rules housed in a relational database; this approach provides for greater flexibility and reconfigurability of the control system. A key result that is demonstrated with this solution is that effective predictive maintenance can result from focusing on data consolidation, resource coordination and flexibility, while utilizing straight-forward prediction algorithms.