What We Publish

On this page, you can find links to our internal MSDL report publications, as well as abstracts and bibliographic information for publications in peer-reviewed journal articles and conferences. Older publications by Dr. Collette before he founded the MSDL are also included. Where allowed by copyright, full text downloaded or pre-prints of articles are also provided. Papers where source code and data is provided for provenance will have a link below the abstract to access the code or data. Zotero metadata is also provided for the publications. 


Department of Naval Architecture 
and Marine Engineering
210 Naval Architecture and Marine Engineering
2600 Draper Dr., Ann Arbor MI 48109
Linkedin picture


Zhu, J., M. Groden, and M. Collette. 2013. "Bayesian Updating of Marine Structural Reliability Models based on In-Service Measurements". The 6th International Conference on Structural Health Monitoring of Intelligent Infrastructure. Hong Kong 9-11 December.

Estimating future reliability and maintenance needs of aging ship structures is a growing challenge. It is attractive to use structural health monitoring data to forecast future structural conditions via a model updating approach. A unique challenge in model updating for ships is their inherent mobility, which subjects them to different environments and operational profiles at different points in their service lives. Thus, updating and forecasting techniques must not only be able to track the type of usage experienced to date, but also potentially forecast future performance in different conditions. This paper explores a model updating approach concentrating on structural fatigue failures, examining both seaway load and fatigue capacity updating. In seaway load analysis, the loads experienced by the vessel are calculated in a series of operational cells corresponding to unique combinations of sea state, heading, and speed. Lifetime loading profiles are then constructed through probabilistic integration of these individual cells. A cell-based updating strategy is presented where Markov Chain Monte Carlo methods are used to approximate model bias and uncertainty factors by comparing predicted loads to experienced loads in cells observed to date. The errors are estimated and used to forecast updated loads in unobserved cells, leading to a refined prediction of lifetime loading. The fatigue capacity updating model uses a lognormal modelling approach to calculate and analyze the probability of crack initiation time and associated reliability. This model is extended to via efficient formulas for forecasting the expected number of fatigue cracks over time in grillage-type structures with multiple similar fatigue-prone details. A Bayesian network is used to develop revised capacity parameter estimates and estimates of future rates of fatigue cracking through an inference approach-based method on observed failures and updated loading data. Examples are provided to show the updating power of this approach and a simple method of combining the two updating approaches.

Mark Groden and Matthew Collette. "Bayesian Updating of Marine Structural Reliability Models Based on In-Service Measurements." 2013 SNAME Annual Meeting. Nov 6-8 2013. Bellevue, WA

Estimating future reliability and maintenance of aging ship structures is a growing challenge. Our improved ability to sense structural loads and responses through structural health monitoring systems has resulted in an increased awareness of how structures are being used. An attractive use of this data is the forecasting of future structural conditions. This research explores a model updating approach concentrating on structural fatigue failures and observed permanent set, examining fatigue capacity updating. A probabilistic S-N fatigue crack initiation model is presented. By using a lognormal modeling approach, the probability of time to crack initiation can be calculated analytically. This model is extended via efficient formulas for forecasting the expected number of fatigue cracks over time in grillage-type structures with multiple similar fatigue-prone details. The updating power of this network is extended via a permanent set model relating load and permanent set assuming uniform pressure. Coupling this permanent set model with the fatigue model, through local stress, is demonstrated to provide a more accurate prognosis than the fatigue model alone. The parameters of this model are represented using a Bayesian Network (BN). The BN is used to develop revised capacity parameter estimates and estimates of future rates of cracking through an inference approach based on observed failures. The model is tested against synthetic data for grillages with varying numbers of fatigue-prone connections. Using the log-normal model, Monte Carlo simulations are run to generate cracking and permanent set histories. A Bayesian network is used to develop revised capacity parameter estimates and estimates of future rates of fatigue cracking through an inference approach based upon observed failures and updated permanent set data. The utility of fusing multiple measurements to update the future prognosis is evaluated.

Temple, D. and M. Collette. 2013. "Optimum Lifetime Maintenance Schedule for Naval Vessels Subject to Fatigue and Corrosion", 12th International Symposium of Practical Design of Ships and Other Floating Structures (PRADS 2013), 1:395-402. Changwon, Korea.

As current naval fleets age, and budgetary issues force governments to extend the service life of their ships long past the point they were originally designed for, maintenance costs over the lifetime of vessels are becoming an increasingly large burden. Maintenance schedules for most naval vessels are largely based on their availability, usually calling for repairs on fixed intervals that are not chosen with the specifics of the ship's structure in mind. Because of this it is desirable to optimize the scheduling of a vessel's maintenance cycles for the design of the physical structure of the ship. A framework is presented to schedule maintenance cycles for naval vessels in order to minimize the lifetime costs for the ship. Using this model and the structure of a notational DTMB-5145 hull form the maintenance schedule for the vessel is optimized in order to minimize the lifetime costs.

Devine T. and M. Collette. 2013. "Application of the Bayesian Optimization Algorithm to Naval Structural Design", 12th International Symposium on Practical Design of Ship and Other Floating Structures (PRADS 2013), 2:694-701. Changwon, Korea.

As hull complexity of modern surface combatants grows, designers require new methods to quickly and effectively search a design space in early stage design. It is essential that designers can, with minimum effort, evaluate the worthiness of potential design solutions. In situations where such solutions are non-intuitive, the optimization routine becomes the primary means to explore the design space. Evolutionary Algorithms (EAs) and particularly Bayesian Optimization Algorithms (BOAs) shows promise as a useful methodology for accomplishing this task. In order to demonstrate the BOA’s effectiveness, the U.S. Navy’s aluminum hull T-Craft is used as a design case. A hypothetical midship section is evaluated and optimized by the BOA for production cost under a given set of structural strength constraints. The results are contrasted with a single objective genetic algorithm (SOGA), a popular current alternative used in the field of design, comparing both the effort or fitness calls used to reach the solution and the resulting solution itself.

Zhu, J. and M. Collette. 2013. "Reliability Based Structural Failure Modeling through Bayesian Networks", 12th International Symposium on Practical Design of Ships and Other Floating Structures (PRADS 2013), 2:739-746. Changwon, Korea.

Structural fatigue reliability analysis for marine and offshore structures is subjected to a significant amount of uncertainty. Through inspection and monitoring, information such as crack sizes can be obtained and used to update reliability estimates. Fatigue cracking in a stiffened panel has been modeled here by a dynamic Bayesian network (BN) which allows updating of reliability estimations with life cycle data. Most of the present BN inference methods attempt to model continuous random variables by numerical approximations using static discretization at each network node. However, such approaches are computationally inefficient for low probability of failure events. A special discretization scheme is developed for such reliability inference. The proposed discretization scheme will focus on how to evaluate the tail region probability accurately instead of the whole distribution shape. An iterative algorithm has been applied which involves dynamically partitioning the intervals at each iteration according to the posterior distribution of each node. This proposed framework is compared with static discretization using the synthetic crack growth data simulated from extended finite element method (XFEM).

Nappi, N. and M. Collette. 2013 "Structural Design of Naval Vessels: Recent Developments and Emerging Challenges", ASNE Day 2013, Arlington, VA.

This paper presents an overview of current naval structural design procedures and challenges, as well as areas for future exploration. Unique features of the design, production, and operation of naval ship structures are reviewed. Then a two-task breakdown of the structural design process into achieving fitness-for-purpose, and achieving higher design goals is presented. While these tasks are not independent, past developments have tended to focus more heavily on one task or the other. The current state-of-the-art and recent developments such as goal-based standards, classification of naval vessels, and complexity estimation are reviewed. Novel structural configurations such as the advanced double-hull design are reviewed as part of the focus on design for production. Significant opportunities for future development in terms of regulatory approaches, design frameworks, optimization, and design approaches are presented.

Collette, M. and R. Sielski. 2013 "Aluminum Ship Structures", ASNE Day 2013, Arlington, VA.

The U.S. Navy’s use of aluminum for vessel structures has been growing in recent years, with several classes of ships using aluminum as a structural material for their hulls or superstructure. Aluminum provides a significant weight advantage at a slightly increased cost over a comparable steel structure, which has advantages for high-speed craft and weight or stability restricted vessels. However, recent experience has underlined that aluminum is not simply a lighter version of steel, and constructing ships without considering aluminum’s unique properties can lead to extensive maintenance issues in service. This paper reviews recent work on aluminum, covering material issues including new alloys, sensitization and stress corrosion cracking, tensile and compressive limit state prediction, response under lateral load, fatigue initiation, crack propagation, and wider issues of design and lifecycle analysis. Future research and development needs are then presented based on the state-of-the-art.

Collette, M. and J. Lynch. 2013 "Lifecycle Support for Naval Ships Based on Strucutral Health Monitoring: Data to Decision Strategies", ASNE Day 2013, Arlington, VA.

There is growing interest in extending the service life of existing naval vessels. For the structural components of the hull, supporting such extensions is presently difficult as the vessel’s loads are continuously changing as the vessel’s war-fighting capabilities and operational profile of the vessel are typically evolving over its life-span. To best manage this challenging situation, lifecycle support systems are direly needed to predict the health of the hull, estimate remaining life in hull components and to estimate the costs of hull repairs over the anticipated remaining service life of the vessel. Towards this end, various enabling technologies are now emerging that have the potential to usher in powerful lifecycle decision support systems for the U.S. Navy. For example, sensing technologies have dramatically improved in recent years with dense instrumentation of in-service hulls possible. However, sensors and the data they generate are only useful if they are integrated into a comprehensive data-to-decision (D2D) framework that provides port engineers and ship owners with valuable information derived from sensor data that quantitatively supports their decision making process. This paper proposes a comprehensive D2D framework for the naval engineering community based on the integration of sensor networks, database systems, and computational tools, all working together to interrogate raw sensor data for extraction of pertinent ship lifecycle information required for optimal decision making.

M. Nelson, D.W. Temple, J.T. Hwang, Y.L. Young, J.R.R.A. Martins, M. Collette Simultaneous Optimization of Propeller–Hull Systems to Minimize Lifetime Fuel Consumption. Applied Ocean Research, Volume 43, October 2013, Pages 46–52

In traditional naval architecture design methodologies optimization of the hull and propeller are done in two separate phases. This sequential approach can lead to designs that have sub-optimal fuel consumption and, thus, higher operational costs. This work presents a method to optimize the propeller–hull system simultaneously in order to design a vessel to have minimal fuel consumption. The optimization uses a probabilistic mission profile, propeller–hull interaction, and engine information to determine the coupled system with minimum fuel cost over its operational life. The design approach is tested on a KCS SIMMAN container ship using B-series propeller data and is shown to reduce fuel consumption compared to an optimized traditional design approach.

Temple, D., and M. Collette. 2013 (Accepted Nov 2012). "Optimization of Structural Design to Minimize Lifecycle Maintenance Cost", MARSTRUCT 2013, Aalto, Finland.p 525-532, 2013

In the design of most naval combatants, the internal structure is designed to minimize the overall weight of the vessel while meeting certain constraints introduced by minimum strength requirements or regulations and standards. This approach to design, however, does not take into account the maintenance that will be necessary over the course of the vessel's life and the costs that these repairs can cause the ship's owner to incur. This work assesses year-by-year costs of the naval vessel over her service life to determine a life-cycle maintenance cost value for the ship. The model captures the logistical aspects of structural repair costs by utilizing a maintenance schedule over the vessel's life. This model is integrated with a multi-objective genetic algorithm to find the trade-offs between the structural weight and life-cycle maintenance costs.

Rigterink, Douglas; Collette, Matthew; Singer, David J. A method for comparing panel complexity to traditional material and production cost estimating techniques. Ocean Engineering Volume 70 September 15, 2013. p 61-71.

Abstract In this paper a metric for assessing the producibility of a stiffened grillage structure in the early stage design is presented. This metric is comprised of seven separate producibility drivers that are dependent on the properties of the panel. The new producibility metric is compared with a traditional costing method in a two-objective genetic algorithm to show that there is a competition between the two methods, meaning that the new metric gives a designer additional information at the early design stage. The estimated costs versus producibility scores for several grillage sizes show that large gains in producibility can be made without large increases in estimated costs up until a critical point where a knee in the Pareto front occurs. Furthermore, it was found that the individual producibility elements are highly dependent on the size of the panel tested and the constraints the panel must meet. In general, producibility is a key factor to be considered in structural design and should be accounted for in any structural optimization.

2012 unselected
2011 unselected
2010 unselected
2009_2007 unselected
2006_2003 unselected