Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
ISSN 1759-3158
CSETS: 19
TRENDS IN COMPUTATIONAL STRUCTURES TECHNOLOGY
Edited by: B.H.V. Topping, M. Papadrakakis
Chapter 3

Optimization of Life-Cycle Maintenance Strategies under Uncertainties: Role of Inspections

D.M. Frangopol1 and L.A.C. Neves2

1Department of Civil and Environmental Engineering, ATLSS Center, Lehigh University, Bethlehem PA, United States of America
2Department of Civil Engineering, New University of Lisbon, Quinta da Torrre, Caparica, Portugal

Full Bibliographic Reference for this chapter
D.M. Frangopol, L.A.C. Neves, "Optimization of Life-Cycle Maintenance Strategies under Uncertainties: Role of Inspections", in B.H.V. Topping, M. Papadrakakis, (Editors), "Trends in Computational Structures Technology", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 3, pp 55-74, 2008. doi:10.4203/csets.19.3
Keywords: bridges, bridge management, cost, deterioration, existing structures, life-cycle performance, maintenance, optimization, simulation, uncertainty, updating.

Summary
The deterioration of structures is an inevitable consequence of time. In large structure networks, the cost associated with deterioration is a significant part of available budget. As a consequence, significant investments have been made in the development of decision aiding tools that help bridge managers in defining optimal maintenance strategies. Bridge management systems (BMS) were introduced in the 1980's in the United States [1,2], based on statistical inference on long record of visual inspection results. In the present paper, the advantages and limitations of these systems are presented and discussed. Recent developments introduced by the authors and co-workers are also presented and discussed. First, a continuous deterioration model, based on survivor functions is defined [3]. Performance is assessed in terms of the probability of finding a defect, defined in a manner similar to that employed in existing BMS. The simplicity of the model allows the integration with optimization tools in a fast and straightforward manner, allowing the analysis of redundant systems.

The second model discussed was introduced by Frangopol [4]. In this model, performance is defined as the reliability index of a structure, such as a bridge. The deterioration of performance is defined as a bi-linear function, associated with a period of initiation of deterioration followed by a period of constant deterioration rate of the reliability index. The maintenance model is defined by eight parameters. The uncertainty associated with the deterioration process and the effects and times of application of maintenance actions are taken into account by defining all these parameters as random variables. The use of the reliability index as a measure of performance results in a more consistent approach.

In order to capture the advantages of considering performance in terms of a visual inspection result and in terms of a safety analysis, the authors [5] proposed a model defining performance in terms of the condition index, the safety index, and the cumulative maintenance cost. The model defines the lifecycle condition and safety indices in a manner similar to that defined by Frangopol [4] for the reliability index. The condition and safety profiles under no maintenance and the effects of each maintenance action are defined by random variables. Deterministic and probabilistic relations between the condition index and the safety index can be defined. In order to find the best possible maintenance strategies for existing bridges, optimization tools can be integrated with these models. A multi-objective optimization tool, based on genetic algorithms was combined with the model proposed in [5], in order to maximize safety, minimize condition (condition index is increasing with deterioration), and minimize lifecycle cost. The results obtained show that a wide range of optimal solutions exist, but, in all cases, the combination of preventive (routine) maintenance actions with essential or repair maintenance actions, results in better overall performance for lower lifecycle costs.

Lastly, a methodology to incorporate the results of visual inspections in the model proposed in [5] is presented. In this model, Bayesian updating techniques are employed to merge two sources of information, both considered uncertain. In fact, both the expert judgment and statistical inference used to define the deterioration model and the effects of maintenance actions, and the results of observations and classification of deterioration of bridges are associated with uncertainty. Observation can provide accurate information on current performance, but only expert judgment and experience can predict the future performance. The results obtained for the examples analyzed show that, even if low quality inspections are considered, the improvement in predicted performance, associated with a reduction in uncertainty is significant, allowing better maintenance decisions.

References
[1]
H. Hawk, E.P. Small, "The BRIDGIT bridge management system", Structural Engineering International, IABSE, 8(4), 303-314, 1998. doi:10.2749/101686698780488758
[2]
P.D. Thompson, "The Pontis bridge management system", Pacific Rim TransTech Conference: International Ties, Management Systems, Propulsion Technology, Strategic Highway Research Program, Seattle, 500-506, 1993.
[3]
S.I. Yang, D.M. Frangopol, L.C. Neves, "Optimum maintenance strategy for deteriorating structures based on lifetime functions", Engineering Structures, 28(2), 196-206, 2006. doi:10.1016/j.engstruct.2005.06.024
[4]
D.M. Frangopol, "A probabilistic model based on eight random variables for preventive maintenance of bridges", Optimum Maintenance Strategies for Different Bridge Types, London, U.K., 1998.
[5]
L.C. Neves, D.M. Frangopol, "Condition, safety and cost profiles for deteriorating structures with emphasis on bridges", Reliability Engineering & System Safety, Elsevier, 89(2), 185-198, 2005. doi:10.1016/j.ress.2004.08.018

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £88 +P&P)