Failure prediction in hierarchical equipment system: spline fitting naval ship failure

Predicting equipment failure is important because it could improve availability and cut down operating budget. Previous literatures have attempted to model failure rate with bathtub-formed function, weibull distribution, Bayesian network, or AHP. But these models perform well with sufficient amount of data and could not incorporate the two salient characteristics; unbalanced category and sharing structure. Hierarchical model has the advantage of partial pooling. The proposed model is based on Bayesian hierarchical B-spline. Time series of failure rate of 99 ROK Naval ships have been modeled as hierarchical model, where each layer corresponds to ship engine, Engine type, and Engine archetype. As a result of analysis, the suggested model predicted the failure rate of an entire life time accurately in multiple situational conditions. The proposed model can contribute greatly to the following areas. First, failure rate prediction could be used as a quantitative reference when establishing a maintenance policy. Proper maintenance not only improves the availability and mission completion rates, but also reduces budget by reducing unnecessary maintenance. Second, we have shown how qualitative knowledge, such as the descendancy or construction era, could be incorporated into the model; this approach was justified by further analyzing the relationship between learned parameters. Lastly, in a more broader perspective, predicted failure trend can be a qualitative reference for designing the optimal life cycle of a ship. For instance, based on our results, failure rate increases dramatically as the ship becomes senile. Therefore optimal retirement period could be decided by balancing the maintenance and construction cost.


Presenter biography:
Hyunji Moon

Hyunji Moon is a senior undergraduate student in the department of industrial engineering Seoul National University. She is interested in analytics and stochastic programming based on Bayesian inference, statistical learning, and optimization. More specifically, she has coauthored two publications in system dynamics simulation and time series domain and was an invited speaker in Stan conference, Cambridge. She is a founder and current CEO of analytics company, Nextopt which has successfully completed consulting projects and hosting its forecasting web service at