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

2015 unselected

David Patrick Hodapp; Co-Chairs: Armin W. Troesch and Matthew D. Collette. "The Use of High-Fidelity Numerical Models in Ship Structural Fatigue Predictions". P.h.D University of Michigan-Ann Arbor. 2014

The advent of high-fidelity, time-domain seakeeping codes over the past several years now permits reasonably accurate numerical simulations of nonlinear ship motions and responses. At present, these codes are primarily used in the prediction of lifetime extreme or design events. However, there is also a need to accurately characterize these nonlinear behaviors when considering fatigue fracture. In contrast to the former, fatigue fracture necessitates a consideration of not just the largest loading cycles, but all loading cycles in the order in which they occur. To appreciate the scope of the problem, one must further consider the non-stationary, stochastic nature of the marine environment, in which ship structures typically experience upwards of 10^8 time-dependent cycles during a nominal service life. For this reason, the marine industry continues to rely almost exclusively on classification rules-based fatigue assessments, centered around linear seakeeping theory and a linear damage hypothesis. This dissertation advances the current state-of-the-art with respect to fatigue fracture in the marine industry by considering three interrelated issues in concert. The first involves the time-dependent nature of the fatigue inducing loads which, given a linear damage hypothesis, is typically a mute point. Intuitively, however, the order of the loading does matter and extreme overloads are not randomly dispersed throughout, but clustered together during physical storms. Therefore, the present work addresses the simulation of long, time-dependent (storm model) stress sequences which are stationary at one timescale (i.e., on the order of hours), yet decidedly nonstationary over longer intervals. Two categories of nonlinearities are simultaneously addressed. They are considered to arise from nonlinear ship motions and responses, and from the conversion of the resultant structural loading to equivalent fatigue damage and/or crack growth. In the present work, the former is taken to include the contribution of nonlinear wave-induced bending and whipping responses, whereas the latter encompasses the material hysteresis or load interactions inherent to variable amplitude loading. In accounting for this material hysteresis, focus is shifted from a hypothetical fatigue damage criterion (i.e., applicable to the crack initiation and early crack growth phases), to macroscopic fatigue crack growth behavior within the context of a damage tolerant design. By considering the time-dependent nature of the fatigue inducing loads, a novel modeling approach is proposed which extends the finite element analysis of plasticity-induced crack closure to variable amplitude, high-cycle fatigue predictions. In doing so, cycle-by-cycle material hysteresis is included through a time-dependent crack “opening” level. This approach is demonstrated to be both consistent and convergent and, in contrast to previous numerical studies of a similar scope which rely on a strip-yield based model, permits the incorporation of a material constitutive model suited to cyclic plasticity in structural steels. Implementing this model within the context of storm model loading, several aspects of the fatigue fracture process are explored. Elucidated behaviors include the influence of nonlinear ship responses, the significance of physical storms, and the random nature of the fatigue process over a finite interval of ship operation.

Jiandao Zhu; Chair: Matthew D. Collette. "Life Cycle Fatigue Management for High-Speed Vessel Using Bayesian Updating Approaches". P.h.D University of Michigan-Ann Arbor. 2014

Structural fatigue cracking in lightweight high-speed vessel structures is a central maintenance and lifecycle costing concern. While traditional pass-fail approaches provide a simple design oriented metric to limit the amount of fatigue cracking observed in service, these approaches struggle to make accurate mid-life predictions of future fatigue performance and the associated uncertainties and risks. A stochastic method of modeling crack growth and fatigue life prediction is proposed based on dynamic Bayesian networks. This is a graphical model represented by sequences of random variables with defined conditional independences between these variables. The aim is not only developing a computationally efficient and robust fatigue life prediction model, but also to incorporate the life cycle monitoring results to determine as-built fatigue properties of vessels via a Bayesian updating approach. A robust discretization technique is also studied to facilitate determining specific reliability levels. The model is then extended to consider variable amplitude loading with a Markov chain Monte Carlo load updating strategy. By sampling from the posterior load distribution data at available sea states, the uncertainties of engineering model used at the initial design stage can be identified and corrected. Thus, a more accurate load prediction integrated with through life information updating is obtained. The proposed framework is then further extended by utilizing simulation data for a specific structural detail generated by using extended finite element method (XFEM). The core idea behind XFEM is to generate the mesh independent of discontinuity domains which makes cycle-by-cycle fatigue cracking simulation possible. The example case study addresses a stiffened panel from joint high-speed sealift (JHSS) with load information simulated by the Large Amplitude Motion Program (LAMP). The results show that both load updating and crack inspection updating are necessary for accurate reliability estimation.

Liu, Y. and M. Collette. 2014. "Surrogate-assisted robust design optimization considering interval-type uncertainty". In Proceedings of the 2nd International Conference on Maritime Technology and Engineering (MARTECH). Page 287-293. Lisbon, Portugal.link_to_journal

This paper presents an efficient surrogate assisted max-min optimization scheme applied in solving a marine structure design problem considering interval-type of uncertainty. Preliminary stage marine structure designs decisions tend to have a large impact on the acquisition cost and future ownership cost. However, in the preliminary design of innovative vessels, much information about the structure remains unknown to the designers. This paper adopted a max-min searching approach to deal with this epistemic type uncertainty in structural design. It is demonstrated that the proposed method is successful in addressing the interval-type uncertainty design problem with limited computational expense.

Zhu, J. and M. Collette. 2014. "Life cycle fatigue management for high speed vessels by integrating structural health monitoring data".  In Proceedings of the 2nd International Conference on Maritime Technology and Engineering (MARTECH). Page 595-603. Lisbon, Portugal.

Further developments in a life cycle fatigue management approach by integrating measured bending moment data and crack inspection records is presented, with a focus on comparing different forms of the lifetime updating model structure. A lifetime load updating strategy is proposed for the lifetime fatigue load distribution via limited load measurements. A dynamic Bayesian network - based stochastic crack growth model is constructed to incorporate crack inspection results and update the fatigue parameters for a more accurate future life prediction. This framework is demonstrated for a cracked stiffened panel on the deck of a joint high-speed sealift ship. The results demonstrate that by incorporating both load data and crack inspection records into the proposed structural health monitoring framework, more accurate crack size prediction and reliability estimation are both achieved. The structure of the lifetime load updating model is shown to have a significant impact on the final results.

Devine, T. and M. Collette. 2014. "Use of Network Metrics with the Bayesian Optimization Algorithm in Marine Structural Design". COMPIT 2014, 365-377, Redworth, United Kingdom.link_to_compit proceedings 2014

Early stage marine design contains high uncertainty while affording designers flexibility to alter the design. An ideal early-stage optimization approach should investigate the design space, while concurrently learning problem structure and variable dependencies. The Bayesian Optimization Algorithm (BOA) applies Bayesian learning in an optimization process to develop a network representing the design variable space. Network analysis is then conducted to determine information about the variables, objective function and constraints shaping the design space. The optimizer is demonstrated on a notional US Navy TCraft midship section. Results are analyzed and discussed to highlight the learning potential of the BOA.

Zhu, J. and M. Collette. 2014. "Updating Structural Engineering Models with In service Data: Approaches and Implications for the Naval Community" ASNE Day 2014, Arlington, VA.link_to_journal

Despite growing computational and sensing power, naval structural design and analysis remains focused minimizing structural system weight while ensuring that estimated stress levels remain below allowable thresholds. While this approach allows for the rapid design of new vessels meeting existing structural requirements, it struggles to fulfill the U.S. Navy’s growing need for lifecycle support of aging assets. To address these new tasks, extensions of current structural analysis and design tools are required. This paper presents a comprehensive framework for managing fatigue cracks on aging assets, extending traditional design-stage approaches with Bayesian network based model updating. A methodology of updating structural load estimates that is able to account for changing operational profiles is presented, allowing future fatigue loading to be predicted with increased confidence. A Dynamic Bayesian Network (DBN) approach is taken to represent the time-varying growth of fatigue cracks, including both shipboard inspection data and the updated loading. A robust reliability formulation is used to predict future crack growth risks based on the DBN formulation and the inspection data to date. An example is presented for a hypothetical sealift vessel. Finally, a discussion of the implications of such models on Navy design practice, operations, and maintenance is presented

Liu, Y. and M. Collette. 2014. "Improving Surrogate-Assisted Variable Fidelity Multi-Objective Optimization Using a Clustering Algorithm." Applied Soft Computing 24 (November): 482-93.link_to_journal

Surrogate-assisted evolutionary optimization has proved to be effective in reducing optimization time, as surrogates, or meta-models can approximate expensive fitness functions in the optimization run. While this is a successful strategy to improve optimization efficiency, challenges arise when constructing surrogate models in higher dimensional function space, where the trade space between multiple conflicting objectives is increasingly complex. This complexity makes it difficult to ensure the accuracy of the surrogates. In this article, a new surrogate management strategy is presented to address this problem. A k-means clustering algorithm is employed to partition model data into local surrogate models. The variable fidelity optimization scheme proposed in the author’s previous work is revised to incorporate this clustering algorithm for surrogate model construction. The applicability of the proposed algorithm is illustrated on six standard test problems. The presented algorithm is also examined in a three-objective stiffened panel optimization design problem to show its superiority in surrogate-assisted multi-objective optimization in higher dimensional objective function space. Performance metrics show that the pro-posed surrogate handling strategy clearly outperforms the single surrogate strategy as the surrogate size increases.

2013 unselected
2011 unselected
2011 unselected
2010 unselected
2009_2007 unselected
2006_2003 unselected