Correlation based anomaly detection and predictive maintenance

Production equipment is often used without a planned maintenance approach. Such a strategy often leads to unplanned downtime due to unexpected failures. Scheduled maintenance frequently replaces components to prevent unexpected equipment downtime, but increases the time associated with machine downtime and maintenance cost. The emergence of Industry 4.0 and intelligent systems has led to increased interest in Predictive Maintenance (PdM) strategies that can reduce downtime and increase the availability (utilization) of manufacturing equipment. Predictive Maintenance (PdM) is a widely used application in recent years. It is a maintenance method that monitors the situation and evaluates the possibility of failure before data problems occur. Monitors the performance and condition of the equipment during normal operation to reduce the possibility of failure. Predictive maintenance also has the potential to develop sustainable applications in production by maximizing the lifetime of the components. It is used in various industries such as production, infrastructure, health and energy to increase productivity and save operating costs. In this thesis, using artificial neural networks instead of commonly used regression methods, the results were observed and their comparison with other regression models in terms of performance was tried to be shown.