Optimization of a proton exchange membrane fuel cell membrane electrode assembly

A computational framework for fuel cell analysis and optimization is presented as an innovative alternative to the time consuming trial-and-error process currently used for fuel cell design. The framework is based on a two-dimensional through-the-channel isothermal, isobaric and single phase membrane electrode assembly (MEA) model. The model input parameters are the manufacturing parameters used to build the MEA: platinum loading, platinum to carbon ratio, electrolyte content and gas diffusion layer porosity. The governing equations of the fuel cell model are solved using Netwon’s algorithm and an adaptive finite element method in order to achieve near quadratic convergence and a mesh independent solution respectively. The analysis module is used to solve the optimization problem of finding the optimal MEA composition for maximizing performance. To solve the optimization problem a gradient-based optimization algorithm is used in conjunction with analytical sensitivities. By using a gradient-based method and analytical sensitivities, the framework presented is capable of solving a complete MEA optimization problem with state-of-the-art electrode models in approximately 30 min, making it a viable alternative for solving large-scale fuel cell problems.     

Computing eigenvalue bounds for iterative subspace matrix methods

A procedure is presented for the computation of bounds to eigenvalues of the generalized hermitian eigenvalue problem and to the standard hermitian eigenvalue problem. This procedure is applicable to iterative subspace eigenvalue methods and to both outer and inner eigenvalues. The Ritz values and their corresponding residual norms, all of which are computable quantities, are needed by the procedure. Knowledge of the exact eigenvalues is not needed by the procedure, but it must be known that the computed Ritz values are isolated from exact eigenvalues outside of the Ritz spectrum and that there are no skipped eigenvalues within the Ritz spectrum range. A multipass refinement procedure is described to compute the bounds for each Ritz value. This procedure requires O(m) effort where m is the subspace dimension for each pass.     

The effects of password length and reference profile size on the performance of a multivariate text-dependent typist verification system

The performance of Napier et al.'s typist verification algorithm (Keyboard user verification: toward an accurate, efficient, and ecologically valid algorithm, International Journal of Human-Computer Studies 43 (1995) 213-222) was assessed in a text-dependent setting. Twenty-nine subjects typed a 17 character password 50 times. False acceptance and false rejection rates were then calculated as the number of repetitions of the password included in the reference profile was increased from 6 to 20 and the number of digraphs from the password included in the verification process was increased from 2 to 16. The performance of the system (12% total error rate) was found to be comparable with the best results reported in other studies using text-dependent algorithms, and substantially better than that reported in studies using a text-independent paradigm with passwords of this length. The relationship between password length and reference profile size was found to conform to an exponential decay function, which accounted for 92% of the variability in verification error rates.     

Summative evaluation of the SINCGARS Tutor

In an earlier set of studies with a different Intelligently Coached Simulation (Orey, M.A., Fan, H., Park, J., Tzeng, S., & Gustafson, K. (1995). Evaluation of Device operator in a context of a coached simulation environment), we found a retention and a transfer problem. We tried to solve these problems while building a new Intelligently Coached Simulation (the SINCGARS Tutor). The solutions to the two problems were to use an interactive conceptual model for the retention problem and we used photographs of the equipment as the visuals of the simulation. We then tested this new tutor with a group of 22 officers who were not only required to know how to operate the SINCGARS radio, but would be responsible for teaching others in their unit when their training was complete. We had one group of officers train on the real equipment, in pairs, with one instructor available for guidance. The other group of 11 learned via the SINCGARS Tutor. The post test was to put an actual radio into operation while being observed by a trained observer. Results indicate that not only did the transfer problem go away, but officers trained on the computer performed the task more accurately both initially on the immediate post test and again on a surprise four-week delayed post test. The SINCGARS Tutor was found to be a very good training solution.     

Equi-affine Invariant Geometry for Shape Analysis

Traditional models of bendable surfaces are based on the exact or approximate invariance to deformations that do not tear or stretch the shape, leaving intact an intrinsic geometry associated with it. These geometries are typically defined using either the shortest path length (geodesic distance), or properties of heat diffusion (diffusion distance) on the surface. Both measures are implicitly derived from the metric induced by the ambient Euclidean space. In this paper, we depart from this restrictive assumption by observing that a different choice of the metric results in a richer set of geometric invariants. We apply equi-affine geometry for analyzing arbitrary shapes with positive Gaussian curvature. The potential of the proposed framework is explored in a range of applications such as shape matching and retrieval, symmetry detection, and computation of Voroni tessellation. We show that in some shape analysis tasks, equi-affine-invariant intrinsic geometries often outperform their Euclidean-based counterparts. We further explore the potential of this metric in facial anthropometry of newborns. We show that intrinsic properties of this homogeneous group are better captured using the equi-affine metric.     

Efficient algorithms for minimum-cost flow problems with piecewise-linear convex costs

We present two efficient algorithms for the minimum-cost flow problem in which arc costs are piecewise-linear and convex. Our algorithms are based on novel algorithms of Orlin, which were developed for the case of linear arc costs. Our first algorithm uses the Edmonds-Karp scaling technique. Its complexity isO(M logU(m+n logM)) for a network withn vertices,m arcs, M linear cost segments, and an upper boundU on the supplies and the capacities. The second algorithm is a strongly polynomial version of the first, and it uses Tardos's idea of contraction. Its complexity isO(M logM(m+n logM)). Both algorithms improve by a factor of at leastM/m the complexity of directly applying existing algorithms to a transformed network in which arc costs are linear.The final stage of this work was performed while Ron Shamir was a visitor at DIMACS (Center for Discrete Mathematics and Theoretical Computer Science), Rutgers University. Supported in part by National Science Foundation Grant NSF-STC88-09648, and by Air Force Grants AFOSR-89-0512 and AFOSR-90-0008.     

Complete forward displacement solutions for a class of three-branch parallel manipulators

Forward displacement solutions are presented for a class of spatial parallel manipulators. In particular, considered are manipulators consisting of a platform supported through passive spherical joints by three branches, each branch having three-revolute joints forming its main arm. Solutions are described for arbitrary main-arm layouts and for all possible cases of redundant (nine, eight, and seven sensors) and non-redundant (six sensors) sensing of branch main-arm joint displacements. It is demonstrated that closed-form forward displacement solutions can be found for all cases of redundant sensing. Furthermore, it is shown that a closed-form solution can be obtained for one of the two possible cases of non-redundant sensing of the main-arm joint displacements. The only case of joint displacement sensing not allowing a closed-form forward displacement solution is two joints sensed per branch, a case that can be expressed as a 16th-order polynomial of a single variable. Due to the importance of having efficient and failure-safe solutions for the forward displacement problem, it is suggested that appropriate redundancy in displacement sensing should be an important consideration in the design of parallel manipulation devices. © 1994 John Wiley & Sons, Inc.     

Algorithmic stability and sanity-check bounds for leave-one-out cross-validation.

In this article we prove sanity-check bounds for the error of the leave-one-out cross-validation estimate of the generalization error: that is, bounds showing that the worst-case error of this estimate is not much worse than that of the training error estimate. The name sanity check refers to the fact that although we often expect the leave-one-out estimate to perform considerably better than the training error estimate, we are here only seeking assurance that its performance will not be considerably worse. Perhaps surprisingly, such assurance has been given only for limited cases in the prior literature on cross-validation. Any nontrivial bound on the error of leave-one-out must rely on some notion of algorithmic stability. Previous bounds relied on the rather strong notion of hypothesis stability, whose application was primarily limited to nearest-neighbor and other local algorithms. Here we introduce the new and weaker notion of error stability and apply it to obtain sanity-check bounds for leave-one-out for other classes of learning algorithms, including training error minimization procedures and Bayesian algorithms. We also provide lower bounds demonstrating the necessity of some form of error stability for proving bounds on the error of the leave-one-out estimate, and the fact that for training error minimization algorithms, in the worst case such bounds must still depend on the Vapnik-Chervonenkis dimension of the hypothesis class.     

Active fault tolerant control for nonlinear systems with simultaneous actuator and sensor faults

The goal of this paper is to describe a novel fault tolerant tracking control (FTTC) strategy based on robust fault estimation and compensation of simultaneous actuator and sensor faults. Within the framework of fault tolerant control (FTC) the challenge is to develop an FTTC design strategy for nonlinear systems to tolerate simultaneous actuator and sensor faults that have bounded first time derivatives. The main contribution of this paper is the proposal of a new architecture based on a combination of actuator and sensor Takagi-Sugeno (T-S) proportional state estimators augmented with proportional and integral feedback (PPI) fault estimators together with a T-S dynamic output feedback control (TSDOFC) capable of time-varying reference tracking. Within this architecture the design freedom for each of the T-S estimators and the control system are available separately with an important consequence on robust L ₂ norm fault estimation and robust L ₂ norm closed-loop tracking performance. The FTTC strategy is illustrated using a nonlinear inverted pendulum example with time-varying tracking of a moving linear position reference.