A novel mechanism to continuously scan field logs and gain real-time feedback

Abstract

Reliability is characteristic of the system which begins during the concept development phase of a product realization process and continuously or iteratively improved, until its end-of-life. Reliability data along with availability and serviceability (RAS) [1] can commonly be retrieved using the system logs through various data mining techniques. The size of the logs for a typical healthcare modality like the Philips Magnetic Resonance (MR) would be of the order of 3-digit megabyte number per day per installed base. Given the humongous size, various clustering techniques as used in big data processing algorithms [2], grind the data to seek the correct results in a timely and efficient fashion. This post-processing step introduces a temporal shift in analyzing the data much after the events have occurred. For the state of affairs that affects reliability and serviceability, it is important that the condition of the deployed systems is notified to actors who can resolve such issues, meeting shrinking timelines demanded by the service level agreements. This would require the log information to be processed directly at the deployment without causing a system performance regression. This paper talks about such a technique that is implemented within the system purview to improve the lead time and thus increase efficiency of the feedback into the research and development (R & D) department.