The sub-project aims to implement a predictive failure detection approach to Borcelik’s maintenance activities on Continuous Galvanization Line by using EFPF solutions. Most critical assets of Continuous Galvanization Line (CGL) are equipped with vibration sensors and analyzers. The goal of the project is to create a central data warehouse, and run predictive outputs with this data, to optimize maintenance activities, and increase uptime. Another goal is to find a cost-effective connectivity approach for medium critical level assets.

​

The condition of the CGL’s critical equipment is followed by a digital platform in Borcelik. The assets are equipped with vibration, current, velocity and temperature sensors. By condition-based maintenance systems, malfunctions with assets could be identified prior to failure but time-based prediction could not be provided by the existing system. Hence, the steel production requires continuous production, short notice detection could not help to avoid unplanned shutdowns. Production lost, quality problems and safety could be considered as results of malfunction on critical assets.

​

In the sub-project, the condition monitoring system is to provide data related to the critical assets. The scores for these features give an understanding of possible problems of the motors and process. Consortium did some offline analysis of the data for use case validation however still this valuable data from assets and sensors is underutilised. This issue is addressed as the challenge of the sub-project.

​

Status quo, the sub-project will collect data; however, the data is not processed to detect condition change and give early warning to the maintenance team. When maintenance team has early warning of condition change detection earlier, they can plan their maintenance activities more effectively.  Right now, around 30 critical assets on the CGL line are ready to send data. Most of this equipment are motors, fans and pumps. This group is equipped with a sensor - analyzer set.

​

However, the total cost of ownership of that connection is high (especially initial investment cost for analyzer-sensor hardware costs). So, the sub-project will also try to find a more cost- effective condition monitoring approach for medium critical assets to have a better insight about the operation. It means another type of sensor and architecture is needed.

 

Another objective is to find the optimum anomaly detection tool for CGL line. EFPF is offering a set of services around anomaly detection. In the project we will integrate these tools and validate theirs performance for selected equipment by historical and real time data. Next to EFPF services, the sub-project aims to also see the performance of a public cloud service’s performance (AWS Lookout for Equipment or Azure ML) for the same cases.

 

With the optimum architecture,  the main objective is to detect a condition change within 2-3 hours for the connected most critical assets to the platform.