Search

The blast wave propagates into the surrounding air when the high-pressure gas pipeline burst. Both temperature and pressure increase very rapidly. The generation of the blast wave and structural dynamic fracture are tightly coupled together during the burst of the high-pressure gas pipeline. However, fracture behavior and characteristics that influence blast waves’ intensity and spatial shape are rarely studied. This paper establishes a numerical model incorporating strain-based failure criteria for pipe material and fluid-structure coupling algorithm. The dynamic crack growth of the pipe and the outer blast wave propagation can be successfully captured in every timestep. The model is validated by comparing the simulated explosion pressure history and peak overpressure outside the pipeline with the experimental results. The blast wave intensity changes and the distribution of overpressure in the jet direction are clarified. Then some critical parameters of the resulting fracture and blast wave are examined, such as the pipe diameter and wall thickness. Specifically, the relationship between pipe fracture and the generated blast field is discussed, providing a deeper understanding of this highly transient and strong fluid-structure interaction problem. The results would benefit the prediction and accident investigation of high-pressure gas pipeline rupture.

In this paper, we conducted an experiment with four human participants whom were asked to follow a robot gripper with unknown motion as close as possible. The results show that human beings resort to a fairly complicated and continuously changing control strategy. We hypothesize that this strategy can be explained by (1) a feedforward (preview) model of the machine’s motion, and further by (2) human being’s uncertainty in this preview. To test (1), we demonstrate that feedforward control can indeed improve the fitting of the model to the experimental data, and that the feedback gain and the preview length vary across subjects. This model, however, does not explain temporally changing human behavior observed during the experiment. To this end, we propose an extension of the human control model where human behavior is influenced by the preview uncertainty. The extended model incorporates a higher-level planner that determines a target state for a short time interval, and a lower-level controller that meets the target through real-time control. The developed model helps predict detailed human behavior during their interactions with robots.

Recent advances in flexible and wireless sensors, soft materials, and additive manufacturing, have stimulated demands for developing intelligent systems that can achieve multidisciplinary objectives (e.g., mechanical strength, thermal conductivity, state and input estimation, controllability, and others). Existing studies often decouple these objectives through sub-system level design, e.g., topology and material design for mechanical and thermal properties, and filter and sensor/actuator design for observability and controllability, assuming that the sub-systems have minimal influences to each others. To investigate the validity of this assumption, we take a unique angle at studying how the topology of the system influences both structural performance (e.g., compliance under static loads) and input observability (e.g., the error in estimating the loads). We reveal a tradeoff between these two objectives and derive the Pareto frontier with respect to the topology. This preliminary result suggests the necessity of a multiobjective formulation for designing intelligent structures, when significant tradeoffs among system objectives exist.

A key challenge in computational material design is to optimize for particular material properties by searching in an often high-dimensional design space of microstructures. A tractable approach to this optimization task is to identify an encoder that maps from microstructures, which are 2D or 3D images, to a lower-dimensional feature space, and a decoder that generates new microstructures based on samples from the feature space. This two-way mapping has been achieved through feature learning, as common features often exist in microstructures from the same material system. Yet existing approaches limit the size of the generated images to that of the training samples, making it less applicable to designing microstructures at an arbitrary scale. This paper proposes a hybrid model that learns both common features and the spatial distributions of them. We show through various material systems that unlike existing reconstruction methods, our method can generate new microstructure samples of arbitrary sizes that are both visually and statistically close to the training samples while preserving local microstructure patterns.

Designers often express their intents (e.g., on product functionalities and semantics) through shape features. Therefore, collecting such “salient” features from existing shapes and learning their associations with design intents will enable efficient design of new shapes. However, the acquisition of saliency knowledge from a large shape collection has not been accomplished. This paper investigates a gamification approach to this end. In addition, we propose to validate a derived saliency map by its corresponding shape recognition accuracy through crowd-sourcing. This allows a comparison across existing and the proposed saliency acquistion and computation methods. Currentl results show that the proposed method achieves statistically similar recognition accuracy to existing saliency data on a standard shape database, indicating that various saliency maps are equally valid according to the proposed saliency definition. Nonetheless, the saliency data obtained through the proposed game consistently produces reasonable viewpoints across shapes, outperforming existing curvature-based and crowdsourcing approaches. The findings from this study could lead to developments of game mechanisms that are more scalable and cost effective at saliency elicitation than existing paid crowdsourcing approaches.

There is evidence that humans can be more efficient than existing algorithms at searching for good solutions in high-dimensional and non-convex design or control spaces, potentially due to our prior knowledge and learning capability. This work attempts to quantify the search strategy of human beings to enhance a Bayesian optimization (BO) algorithm for an optimal design and control problem. We consider the sequence of human solutions as generated from BO, and propose to recover the algorithmic parameters of BO by maximizing the likelihood of the observed solution path. The method is different from inverse reinforcement learning (where an optimal control solution is learned based on human demonstrations) in that the latter requires near-optimal solutions from humans, while we only require the existence of a good search strategy. The method is first verified through simulation studies and then applied to the human solutions crowdsourced through a gamification of the problem under study [1]. We learn BO parameters from a player with a demonstrated good search strategy and show that applying the BO algorithm with these parameters to the game noticeably improves the convergence of the search from using a default BO setting.

We investigate the cost and benefit of crowdsourcing solutions to an NP-complete powertrain design and control problem. Specifically, we cast this optimization problem as an online competition, and received 2391 game plays by 124 anonymous players during the first week from the launch. We compare the performance of human players against that of the Efficient Global Optimization (EGO) algorithm. We show that while only a small portion of human players can outperform the algorithm in long term, players tend to formulate good heuristics early on, from where good solutions can be extracted and used to constrain the solution space. Incorporating this constraint into the search enhances the efficiency of the algorithm, even for problem settings different from the game. These findings indicate that human computation is promising in solving comprehensible and computationally hard optimal design and control problems.Copyright © 2015 by ASME

A hybrid-electric vehicle powertrain architecture consists of single or multiple driving modes, i.e., connection arrangements among engine, motors and vehicle output shaft that determine distribution of power. While most architecture development work to date has focused primarily on passenger cars, interest has been growing in exploring architectures for special-purpose vehicles such as vans or trucks for civilian and military applications, whose weights or payloads can vary significantly during operations. Previous findings show that the optimal architecture can be sensitive to vehicle weight. In this paper we investigate architecture design under a distribution of vehicle weights, using a simulation-based design optimization strategy with nested supervisory optimal control and accounting for powertrain complexity. Results show that an architecture under a single load has significant differences and lower fuel efficiency than an architecture designed to work under a variety of loading scenarios.

Motivated by continued interest within the design community to model design preferences, this paper investigates the question of predicting preferences with particular application to consumer purchase behavior: How can we obtain high prediction accuracy in a consumer preference model using market purchase data? To this end, we employ sparse coding and sparse restricted Boltzmann machines, recent methods from machine learning, to transform the original market data into a sparse and high-dimensional representation. We show that these ‘feature learning’ techniques, which are independent from the preference model itself (e.g., logit model), can complement existing efforts towards high-accuracy preference prediction. Using actual passenger car market data, we achieve significant improvement in prediction accuracy on a binary preference task by properly transforming the original consumer variables and passenger car variables to a sparse and high-dimensional representation.

Conjoint analysis from marketing has been successfully integrated with engineering analysis in design for market systems. The long questionnaires needed for conjoint analysis in relatively complex design decisions can become cumbersome to the human respondents. This paper presents an adaptive questionnaire generation strategy that uses active learning and allows incorporation of engineering knowledge in order to identify efficiently designs with high probability to be optimal. The strategy is based on viewing optimal design as a group identification problem. A running example demonstrates that a good estimation of consumer preference is not always necessary for finding the optimal design and that conjoint analysis could be configured more effectively for the specific purpose of design optimization. Extending the proposed method beyond a homogeneous preference model and noiseless user responses is also discussed.

Crowdsourced evaluation is a promising method for evaluating attributes of design concepts that require human input. One factor in obtaining good evaluations is the ratio of high-ability to low-ability participants within the crowd. In this paper we introduce a Bayesian network model capable of finding participants with high design evaluation ability, so that their evaluations may be weighted more than those of the rest of the crowd. The Bayesian network model also estimates a score of how well each design concept performs with respect to a design attribute without knowledge of the true scores. Monte Carlo simulation studies tested the quality of the estimations on a variety of crowds consisting of participants with different evaluation ability. Results suggest that the Bayesian network model estimates design attribute performance scores much closer to their true values than simply weighting the evaluations from all participants in the crowd equally. This finding holds true even when the group of high ability participants is a small percentage of the entire crowd.Copyright © 2013 by ASME

We examine the problem of eliciting the most preferred designs of a user from a finite set of designs through iterative pairwise comparisons presented to the user. The key challenge is to select proper queries (i.e., presentations of design pairs to the user) in order to minimize the number of queries. Previous work formulated elicitation as a blackbox optimization problem with comparison (binary) outputs, and a heuristic search algorithm similar to Efficient Global Optimization (EGO) was used to solve it. In this paper, we propose a query algorithm that minimizes the expected number of queries directly, assuming that designs are embedded in a known space and user preference is a linear function of design variables. Besides its theoretical foundation, the proposed algorithm shows empirical performance better than the EGO search algorithm in both simulated and real-user experiments. A novel approximation scheme is also introduced to alleviate the scalability issue of the proposed algorithm, making it tractable for a large number of design variables or of candidate designs.

Several hybrid-electric vehicle architectures have been commercialized to serve different categories of vehicles and driving conditions. Such architectures can be optimally controlled by switching among driving modes, namely, the power distribution schemes in their planetary gear (PG) transmissions, in order to operate the vehicle in the most efficient regions of engine and motor maps. This paper proposes a systematic way to identify the optimal architecture for a given vehicle drive cycle, rather than parametrically optimizing one or more pre-selected architectures. An automatic generator of feasible driving modes for a given number of PGs is developed. For a powertrain consisting of one engine, two motors and two PGs, this generator results in 1116 modes. A heuristic search is then proposed to find a near-optimal pair of modes for a given driving cycle and vehicle specification. In a study this process identifies a dual-mode architecture with an 8% improvement in fuel economy compared to a commercially available architecture over a standard drive cycle.

Active learning refers to the mechanism of querying users to accomplish a classification task in machine learning or a conjoint analysis in econometrics with minimum cost. Classification and conjoint analysis have been introduced to design research to automate design feasibility checking and to construct marketing demand models, respectively. In this paper, we review active learning algorithms from computer and marketing science, and establish the mathematical commonality between the two approaches. We compare empirically the performance of active learning and static D-optimal design on simulated classification and conjoint analysis test problems with labelling noise. Results show that active learning outperforms D-optimal design when query size is large or noise is small.Copyright © 2012 by ASME

Prosthetic heart valve is one of the most successful implantable medical devices. However, introducing better performing and longer lasting prosthetic mechanical heart valves (MHV) into clinical use has been slow because predicting the long term performance of a new valve design is difficult. Although significant progresses in many scientific fronts relevant to prosthetic heart valve development have been achieved, we still have an imperfect understanding of host responses to an implantable medical device and incomplete knowledge in associating hemodynamic characteristics of a valve design to clinical performance. Valve designers, frequently need to over design the valve components to ensure structural safety and thus, sacrifice the opportunity to optimize performance. Complications such as infection, thrombus formation, thromboembolic incidents, and hemorrhage associated to the use of prosthetic valves are still reported and valve designers are working hard to eliminate them. Further advancing scientific knowledge in designing and evaluating prosthetic heart valves is of great interest to many Valve designers and manufacturers. Interfacing Industry and Academic research efforts has been thwarted due to predominantly proprietary issues. Considering the benefits of a better performing MHV to the patients, this industry session will bring researchers from various MHV companies and academic institutions to discuss how to share the results of scientific studies more effectively. This will help accelerate new MHV development without compromising the confidentiality of key valve design information. The issue of standardized MHV testing will also be addressed.