Building a Data-Driven Vital Sign Indicator for an Economically Optimized Component Replacement Policy



Hyung-il Ahn Ying Tat Leung Axel Hochstein


In asset-intensive services, a well-known challenge is to maintain high availability of the physical assets while keeping the total maintenance cost low. In applications of high-value machinery such as heavy industrial equipment, a traditional approach is to perform periodic maintenance according to a runtime-based schedule. Most equipment vendors publish a maintenance schedule based on a “standard” or “average” working environment. In addition, it is a common practice that maintenance schedules from equipment vendors are highly conservative in order to reduce in-field failures which gives an adverse perception of a vendor’s reputation. Therefore, such a schedule may not result in satisfactory performance as measured according to the owner’s business objectives. Also, the assumption of normal operating condition may not apply in some situations. For example, stresses due to frequent overloading, continuous usage of engine at a high rate in tough environments, machine usage beyond its designed capacity can serve as good contributors to excessive wear and premature failures. In this paper we propose a novel computational framework to build a data-driven economically optimized vital sign indicator for a given component type and an economic criterion (e.g., average maintenance cost per unit runtime) by combining different sources of historical data such as total runtime hours, load carried, fuel consumed and event information from sensors. This new vital sign indicator can be viewed as a transformed time scale and used to find the optimal threshold value (or “scheduled replacement time equivalent”) for a component replacement policy. Our case study was based on the collected data from 50 mining haul trucks over about 6 years in one of the largest mining service companies in the world. We present that the new vital sign indicator-based replacement policy for a critical component type largely improves on the traditional runtime-based schedule in terms of a given economic criterion, achieving a lower total maintenance cost of the enterprise.

How to Cite

Ahn, H.- il ., Tat Leung, Y. ., & Hochstein, A. . (2014). Building a Data-Driven Vital Sign Indicator for an Economically Optimized Component Replacement Policy. Annual Conference of the PHM Society, 6(1).
Abstract 2 | PDF Downloads 1



Vital Sign Indicator, Component Replacement Policy, Predictive Asset Management, Equipment Health Monitoring, Individualized Cumulative Hazard, Economic Optimization

Bair, E., Hastie, T., Paul, D., & Tibshirani, R. (2006). Prediction by supervised principal components. Journal of the American Statistical Association, 101(473).

Banjevic, D., Jardine, A., Makis, V., & Ennis, M. (2001). A control-limit policy and software for condition-based maintenance optimization. INFOR-OTTAWA-, 39(1), 32–50.

Cristianini, N., & Shawe-Taylor, J. (2000). An introduction to support vector machines and other kernel-based learning methods. Cambridge university press.

Fox, J. (2002). Cox proportional-hazards regression for survival data. See Also.

Hastie, T., Tibshirani, R., Friedman, J., & Franklin, J. (2005). The elements of statistical learning: data mining, inference and prediction. The Mathematical Intelligencer, 27(2), 83–85.

Jardine, A., Banjevic, D., Montgomery, N., & Pak, A. (2008). Repairable system reliability: Recent developments in CBM optimization. International Journal of Performability Engineering, 4(3), 205.

Jardine, A. K., & Tsang, A. H. (2013). Maintenance, replacement, and reliability: theory and applications. CRC press.

Scholkopf, B., & Smola, A. (2002). Learning with kernels. MITpressCambridge.

Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occurrence. Oxford university press.

Therneau, T. M. (2000). Modeling survival data: extending the Cox model. Springer.

Wu, X., & Ryan, S. M. (2011). Optimal replacement in the proportional hazards model with semi-markovian covariate process and continuous monitoring. Reliability, IEEE Transactions on, 60(3), 580–589.

Zaki, M. J. (2001). SPADE: An efficient algorithm for mining frequent sequences. Machine Learning, 42(1-2), 31–60.
Technical Papers