Diagnosing development software release to predict field failures

With the advancement of analytical engines for big data, the healthcare industry has taken a big leap to minimize escalations on healthcare expenditure, while providing a reliably working solution for the customers based on the slice and dice of the collected information. The research and development (R & D) departments of the healthcare players are providing more focus on the stability and the usage of the system in the field. The field studies have created a reliability based feedback loop that has helped R & D provide hotfixes and service packs in shrinking time lines to better answer the customized needs of the user. Given the variety of possible optimizations in the actual usage, the software-hardware product combine such as the Philips Magnetic Resonance (MR) modality has to ensure that the business critical workflows are ever stable. In a nutshell, fault prediction becomes an important aspect for the R & D department because it helps address the situation in an effective and timely fashion, for both the end-user and the manufacturer to alleviate process hiccups and delays in addressing the fault. Reliability growth plot using the Weibull probability plots helps to predict failures that guide reliability centric maintenance strategies ^[1]; however, this will be a passive application of prediction for the new software yet to be released for market. This paper tries to address the case where a fault/failure at the customer-end can be better predicted for software-under-development with the help of analysis of field data. The terms failures and faults are interchangeably used in the paper to represent error events that can occur at an installed base.