Reliability prediction of long-lived linear assets with incomplete failure data

Linear (or continuous) assets such as railway lines, roads, pipelines and cables are essential in modern society. They play an important socioeconomic role and thus need to be maintained in a way that guarantees their reliability. Linear assets usually span long distances and can be divided into different segments, all of which perform the same function but may be subject to different loads and environmental conditions. Linear assets are usually well-engineered, long-lived assets. As a result, when conducting a reliability analysis, we often only have sparse right-censored failure records, as the majority of segments have not yet failed. Furthermore, many existing linear assets were built decades ago, and their failure and maintenance histories have not been recorded in detail until very recently, leading to a left-censored failure dataset. In addition to having incomplete failure data, linear assets, over their life spans, often present multiple failure characteristics and follow mixed failure distributions, typically a combination of exponential and Weibull distributions. A model that can effectively address these three issues has yet to be developed. In this paper, we present a hazard based predictive method to meet this need. Initially, a training hazard dataset is built up based on the available, incomplete historical failure data. This is then used to develop a parametric hybrid empirical hazard model which is finally used for reliability analysis and prediction.