Large-Scale HMMWV Model Parameterization Using an Activity-Guided Stochastic Search Algorithm
Bryon Sohns, Hosam K. Fathy, and Jeffrey L. Stein

The goal of this project is to develop a method to efficiently parameterized high fidelity models of dynamic systems. One way to improve efficiency is to use activity to estimate the importance of different model parameters. Combining this technique with an appropriate optimization algorithm will greatly enhance the efficiency of the identification process. Since this problem is likely to be non-linear and have many local optima, an evolutionary optimization algorithm is an appropriate choice. This algorithm also has guaranteed convergence properties whether or not activity is used. In this presentation an activity-guided random walk optimization algorithm is described and compared to more traditional methods. It is shown that activity does have some relation to the importance of model parameters. Also, activity is combined with a random walk evolutionary algorithm to identify the parameters of a high-fidelity simulation model of a HMMWV. This model has roughly 20 rigid bodies and more than 100 parameters. Results show that the activity guided random walk efficiently converges to the locality of the global optima.

Approximate Similitude: Quantification and Use in Dynamic System Scaling
Burit Kittirungsi, Hosam K. Fathy, and Jeffrey L. Stein

Previous work has developed a generalized scaling framework which allows engineers to scale a proven design to satisfy new desired dynamic characteristics and simultaneously achieve adequate dynamic similitude. Such a framework employs dimensional analysis to derive scaling laws and add activity analysis to it to help discard the scaling of the least important parameters, thereby gaining more flexibility in scaling. Nevertheless, difficulties still remain and arise when the scaling laws identified as the more important ones cannot be followed exactly; thereby precluding absolute similitude. For this reason, this paper develops, for the first time, a generalizable metric which can quantify, on a continuous basis, the closeness of a design to achieving similitude. As a result, this metric allows the use of a multi-objective optimization problem to find design solutions representing trade-offs between design requirement satisfaction and dynamic similitude achievement. The viability of the method is demonstrated through two examples. The first one illustrates the use of the method to obtain optimal designs of a linear two-degree-of-freedom mass spring damper system. The second example highlights the applicability of the method to a fuel cell's nonlinear air supply system subjected to a set of design physical constraints.

Structural Reduction of Dynamic System Models with Application to HMMWV Multibody Dynamic Model Reduction
Tulga Ersal, Hosam K. Fathy, and Jeffrey L. Stein

The literature deems a model "proper", if it is only as complex as necessary to fulfill its purpose. Ensuring the properness of models is critical to efficient and successful design, analysis, and control of engineering systems, but it requires much time and expertise as systems become more complex, as exemplified by modern automobiles. Thus, there is a growing need for proper-modeling tools for complex systems. This talk will present a new method for reducing overly complex models to render them proper. This method has the following key characteristics: (1) It is energy-based and hence applicable to multidomain models. (2) It takes into account, for the first time, the correlations between the energy-flow patterns in the model. (3) It allows for not only order reduction, but also structural reduction. In mechanical domain, for example, this corresponds to a simultaneous reduction in dynamics and kinematics. (4) It does not require any state transformation, thereby preserving the physical meaning of the model. The modeling and reduction of a HMMWV will be presented to highlight the benefits of the proposed method.

Stochastic Modeling of Terrain Profile and Soil Characteristics
Mr. Lin Li, Dr. Corina Sandu, Dr. Alexander Reid, Mr. Dave Glemming

One fundamental difficulty in understanding the physics of the off-road traction and in predicting vehicle performance is the variability of the terrain profile and soil parameters. These operating conditions are uniquely defined at a given spatial location and a given time. It is not practically feasible to measure them at a sufficiently large number of points to be able to accurately represent the terrain in models. Soil parameters vary substantially from one type of soil to another. Moreover, for the same soil type, the parameters change with environmental conditions difficult to predict, for example the moisture content. This renders traditional analysis tools insufficient when dealing with rough terrain. In this study, mathematical tools to quantify the impact of uncertainties in terrain profile and soil behaviors on vehicle mobility are developed. A polynomial chaos approach is used to reconstruct one-dimensional (along longitudinal direction) stationary and non-stationary terrain profiles. Also, an efficient mathematical method based on the Karhunen-Loeve expansion is developed to reconstruct two-dimensional (along longitudinal and lateral direction) terrain profiles. To investigate the influence of parametric uncertainty of soil on vehicle performance, the polynomial chaos approach is used to establish stochastic shear stress-shear displacement relationship considering randomness in soil moisture, which can be combined with the stochastic pressure-sinkage relationship to explore the vehicle mobility on deformable soil and terrain.

Advanced Semi-Active Control Methods for HMMWV Primary Suspensions
Steve Southward, Mehdi Ahmadian

One promising means for achieving significant performance enhancement of a semi-active suspension control system is to incorporate an accurate non-linear inverse MR damper model in the controller.  This strategy can provide a benefit to the Army for their existing semi-active control algorithms.  Typically, MR dampers are characterized with signals such as DC current and sinusoidal relative velocity inputs, but these signals are not representative of a practical implementation.  Although the resulting models or "maps" are relatively easy to invert, they are deficient for semi-active controller design.  This deficiency can negatively impact the achievable performance.  Experimental response data has been acquired on a COTS MR damper where band limited uncorrelated current and relative velocity signals are simultaneously input to the damper, representing realistic excitations.  Direct comparison with a common static MR damper model clearly indicates the dynamic nature of the actual damper, and the deficiency of the static model.  An adaptive real-time algorithm has been proposed for direct identification of a non-linear dynamic inverse model to empirically fit the MR damper response data.  Future work will include dynamic characterization of an MR damper using the proposed algorithm, followed by a demonstration of semi-active performance enhancement on an actual HMMWV suspension mounted in a quarter-car rig.

Advanced Cooling Strategies for Internal Combustion Engines – Concept, Configurations, and Power Savings
T. Mitchell, M. Salah, J. Wagner, and D. Dawson

Advanced thermal management systems for internal combustion engines can reduce parasitic losses and improve coolant temperature regulation, power consumption, and emissions by better regulating the combustion process with multiple computer controlled electro-mechanical components. The traditional wax-based thermostat valve, coolant pump, and clutch-driven radiator fan are upgraded with servomotor driven actuators. Advanced automotive thermal management systems use computer controllable electric or hydraulic actuators that must work in harmony so that desired thermal conditions can be accomplished in a power efficient manner. A comprehensive nonlinear control architecture has been developed for transient temperature tracking. In addition, four different thermostat configurations have been analyzed to investigate warm-up behaviors and thermostat valve operations. The cooling system configurations considered include factory, two-way valve, three-way valve, and no valve. An experimental system has been fabricated and assembled which features a variable position smart valve, adjustable speed electric coolant pump, variable speed electric radiator fan, 6.0L stationary engine block, and various sensors. In the configured system, the steam-based heat exchanger emulates the heat generated by the engine’s combustion process. Representative numerical and experimental results are discussed to demonstrate the functionality of the thermal management system in accurately tracking prescribed temperature profiles and minimizing electrical power consumption, as well as to examine the effectiveness of each valve configuration in the engine cooling system.

Modeling, Configuration Design, and Control Optimization of the Power-split Hybrid HMMWV
Jinming Liu and Huei Peng

Interest in the pursuit for fuel-efficient technology runs very high in recent years due to the slow-but-sure depletion of crude oil supply, sudden rise in gasoline price, and continued environmental concerns. Hybrid electric propulsion seems to be one of the most promising technologies and is the only one that has been tested successfully on the commercial market. Among its many possible configurations, the power-split (i.e. parallel/series) configuration offers great design and control flexibility and achieves higher overall efficiency. Therefore, it gradually gains popularity in recent years. For example, almost all hybrid vehicles already available or are soon to become available from GM, Toyota, Lexus, and Ford, are of the power-split type. Many power-split hybrid vehicle designs have been suggested in recent development in the automotive industry. How to find the optimal control solution for them and with which to evaluate and compare these designs becomes a valuable research topic. In this research, a virtual power-split super HMMWV is constructed with optional powertrain designs to be analyzed. We will demonstrate a systematic approach to assess these different powertrain configurations by using dynamic programming. Then an optimal control solution is generated for the selected design candidate by using stochastic dynamic programming.

Parameterization Of Fuel Cell Dynamics Using Stack-Level Measurements
Denise McKay, Anna Stefanopoulou

Water management within proton exchange membrane fuel cell stacks is critical for optimizing cell performance. With adequate models for estimating the reactant and water dynamics within the fuel cell electrodes, the model must be parameterized and experimentally validated. Thus, an adequate performance metric must be selected which enables Real Time implementation of the fuel cell model. With estimations of the cell voltage, the performance variable most often selected, we focus on the voltage statistics for a multi-cell stack and the related tradeoffs associated with tuning a model based on the minimum, maximum, mean and median voltages. The variability in the voltage measurement is significantly larger during electrode flooding conditions, as compared to non-flooding conditions. This voltage variability presents a fundamental issue in parameterization and representation of flooding, where we have the largest measurement uncertainty. A combination of deterministic distributed parameter and stochastic models need to be developed to handle the representation of this behavior. This highly-critical and highly-uncertain operation is a ubiquitous characteristic of many systems of technological importance.

Multi-Wheel Vehicle Dynamics and Performance: Wheel Power Distribution and Driveline System Design
Vladimir V. Vantsevich

Dynamics and performance of vehicles with four or more drive wheels, multi-wheel drive vehicles, result from vehicle operational properties such as energy/fuel efficiency, mobility, tractive and velocity properties, vehicle turnability, and stability of motion and handling. No doubt, these vehicle operational properties considerably depend on total power applied to all the drive wheels. At the same time, the vehicle performance strongly depends on the distribution of total power among the drive wheels. In the given road/off-road conditions, the same vehicle with a constant total power at the drive wheels, but with different power distributions among the drive axles and left and right wheels of each axle will perform differently - the criteria of the above-listed operational properties will have different quantities. Power distribution to the drive wheels is largely determined by the vehicle driveline system, which is defined as a part of powertrain located after transmission. A driveline system includes a set of mechanisms and subsystems installed in power-dividing units (PDU). These units include transfer cases, interaxle reduction gears, and drive axles. A list of those mechanisms and subsystems may be formed by limited slip differentials with various torques biases, open and mechanically/electronically lockable differentials, NoSPINs, viscous clutches, torque management/vectoring devices, and many others. This presentation introduces analytical and experimental methods and achievements in designing mechanical and mechatronic driveline systems. The approach to designing driveline systems is that characteristics of a driveline system and a set of power dividing units are established through the inverse vehicle dynamics and vehicle performance analysis and optimization.

Fuel Cell Hybrid Powertrain Optimization for Plug-In Electrical Power Generation
Scott Moura, Dongsuk Kum, Huei Peng, Panos Papalambros, Hosam K. Fathy

Combined plant/control optimization is applied to a PEM hybrid fuel cell vehicle (HFCV) for vehicle to grid (V2G) applications. The HFCV model is developed from past control-oriented models. For the purposes of design optimization, three components (fuel cell stack, compressor, and battery) are made scalable. To construct a control scheme suitable for combined plant/control design optimization, a rule-based method is selected and framed in a manner such that several key parameters are formulated as design variables. Simulation based computations of the objective function are characterized by noise, and therefore inappropriate for gradient-based optimization algorithms. A surrogate modeling method is suggested using neural networks to approximate the physical model. Using the surrogate model, the combined design and controller HFCV model is optimized for maximum fuel economy for a given stationary power demand cycle. The solution is analyzed with respect to various optimality properties, such as constraint activity, Lagrange multipliers, interior & bounded solutions, and varying starting points. The trade-offs between optimal design solutions and constraints is observed and analyzed to analyze optimal design solutions for a PEM HFCV operating as an energy source to the power grid. Multi-objective optimization problems are formulated through parametric studies to elucidate trade offs between different design objectives. A resultant set of “design rules” are formulated to provide a physical engineering interpretation of the conclusions found.

Probabilistic Online Estimation of On-road Vehicle Inertial Parameters
Dongsoo Kang, Hosam K. Fathy, and Jeffrey L. Stein

The performance of active safety systems, especially in SUVs and trucks, depends strongly on uncertain loading conditions that cause significant discrepancies between real and modeled vehicles. In order to function well, they must be calibrated differently for different loading conditions. Calibrating active safety systems to worst-case loading conditions ensures safety over a wide range of loads but visibly sacrifices maneuverability. Alternatively, adapting safety systems to real loading conditions can enhance safety without unnecessarily penalizing maneuverability, but this necessitates online vehicle inertia estimations. Two specific critical vehicle inertial parameters that should be identified as part of such adaptation are vehicle mass and center of gravity height. The proposed vehicle inertial parameter estimation method is equipped with two supervisory processes: reduced Recursive Least Squares (RLS) parametric model through band-pass filtering, and conservative parameter error calculation through the probabilistic analysis. The RLS parametric model can be reconstructed with only a targeted vehicle parameter using minimum sensor packages under relatively high frequency conditions. The online estimated parameter is coupled with conservative upper and lower bounds on this estimate in the 3s sense. The proposed algorithm provides 99% conservative bounds of the parameter as well as the estimate value, and therefore enhances parameter reliability. A simulation example for a large SUV with additional rooftop loading is presented.