J. Mech. Des. 2018; 141(2):021704-021704-9.
Automated Design of Energy Efficient Control Strategies for building clusters using reinforcement learning
Philip Odonkor; Kemper Lewis
J. Mech. Des. 2018; 141(2):021704-021704-9.
From smartphones to electric cars, lithium-ion batteries allow us power our favorite devices. When used in our homes, they allow us to store cheap electricity for later use. These batteries, however, can be prohibitively expensive. But what if a single battery can intelligently be shared by multiple homes? In this paper, we demonstrate the feasibility of this idea by developing an algorithm to autonomously learn the consumption behaviors of multiple real-world homes. This insight is leveraged to produce use strategies, allowing multiple homes to simultaneously enjoy the benefits of a single battery system.
For the Full Paper please visit ASME's Digital Collection.
Co-design is the integrated optimization of the physical plant and controller for an engineering system. The challenge in co-design is determining both time-invariant (physical design) variables and time-variant (control) variables. In co-design, as the size of the problem (number of variables) becomes large, the problem can become too difficult for an all-at-once solution. Our approach extends earlier research by creating a class of multi-subsystem co-design problems where both design and control are formulated and solved. A scalable test problem is used for comparing the proposed decentralized co-design optimization approach against a centralized approach. Results of this study show that the computational time of the proposed decentralized approach increases approximately linearly with respect to an increase in the number of subsystems (variables), while the computational cost of the centralized approach increases nonlinearly.
Sheng Yang and Yaoyao Fiona Zhao
J. Mech Des 140(3):031702-031702-12. doi:10.1115/1.4038922
Part count reduction (PCR) is one motivation for using additive manufacturing (AM) processes. PCR helps simplify product structure, eliminate auxiliary connecters, and reduce assembly difficulties and cost. However, PCR may also increase manufacturing difficulty and the irreplaceability of failed subcomponents. This paper presents a pioneering investigation of how AM-enabled PCR (AM-PCR) impacts lifecycle activities. A new set of design rules and principles are proposed for PCR that lead to lowered cost and enhanced performance. The PCR problem is formulated as a combinatory optimization problem where the objective is minimizing lifecycle cost/performance ratio while ensuring conformance to all constraints (e.g. manufacturing, maintenance, and recycling). To address the challenge of computational cost, a dual-level screening and refinement product redesign framework is presented that first searches for the minimum grouping solution and then refines the remaining combinations using design optimization. This approach will help designers automate the part count reduction process enabled by additive manufacturing while exploring new design innovation opportunities.
For the full article please visit ASME's Digital Collection.
Cityplot is a new visualization technique for engineering design that uses a dimensionally-reduced representation of the design decisions to represent the mapping from the decisions to the criteria upon which a design is judged. The shown Cityplot depicts possible CubeSat constellations to support the 2007 Earth Science Decadal Survey. Each constellation is comprised of up to 4 CubeSats and each CubeSat can select from a list of 7 instruments. Possible CubeSat constellations are “cities” and are placed in a 2d space to be visualized. An individual constellation can also be seen as a table of instruments (rows) being present (black) on a given CubeSat (columns). The benefits, costs and risks of each possible constellation are represented as color-coded “buildings” in each “city”. The criteria in this example are: a tiered count of satisfied Decadal objectives (blue), the average CubeSat Technology Readiness Level (red), lifecycle cost (green), maximum number of lost instruments upon loss of a single satellite (black). A taller building indicates the possible constellation performs better in that criteria. Dark purple “roads” between two designs indicate that only one instrument is either added to or removed from one CubeSat to make one constellation identical to the other. Cityplot simultaneously shows sensitivity of criteria to decisions, criteria tradeoffs and design families via a quick intuitive view of the design space.
The synthesis of functional molecular mechanisms is constrained by the notorious difficulties in fabricating nano-links of prescribed shapes and sizes. Thus, the classical mechanism synthesis methods, which assume the ability to manufacture any designed links, cannot provide a systematic process for designing molecular mechanisms. We propose a new approach to build functional mechanisms with prescribed mobility by only using elements from a predefined "link soup". The resulting synthesis procedure is the first of its kind that is capable of systematically synthesizing functional linkages with prescribed mobility constructed from a soup of primitive entities. Furthermore, the proposed systematic approach outputs the ATLAS of candidate mechanisms, which can be further processed for downstream applications. Although the scope of this technique is rather general, its immediate application is the design of molecular machines assembled from nano-links that either exist in nature or can be fabricated.
For the Full Research Paper see ASME's Digital Collection.
Darren J. Hartl, Edgar Galvan, Richard J. Malak and Jeffrey W. Baur
J. Mech. Des 138(3), 031402; doi: 10.1115/1.4032268
This work is the first demonstration of the use of a parameterized approach to design optimization on a complex engineering problem. Parameterized optimization solves a family of optimization problems as a function of exogenous parameters. When applied to a subsystem of interest, it results in general knowledge about the capabilities of the subsystem rather than a restrictive point solution. The motivation is that it often is necessary to advance the development of a subsystem independent of system-level specifics. This is true during initial research and development efforts or in sensitive military and competitive industrial design environments in which compartmentalization of information is common and necessary. It also is important in systems development projects when the need for concurrency often requires subsystem designers to make progress in the absence of full information about other interfacing subsystems. We solve this specialized design problem using the predictive parameterized Pareto genetic algorithm (P3GA). The approach is demonstrated for the multifunctional design of a structurally-integrated liquid metal circuit intended to provide integrated cooling functionality. A family of optimal design solutions associated with values of external parameters (bounded real numbers) is computed efficiently using P3GA. The demonstration employs both high- and low-fidelity multi-physical engineering models seamlessly and results in general knowledge about the subsystem as a function of parameters associated with other interfacing subsystems.
For the full paper please see ASME's Digital Collection
Kurt Maute, Anton Tkachuk, Jiangtao Wu, H. Jerry Qi, Zhen Ding and Martin L. Dunn
J. Mech. Des 137(11), 111402; doi: 10.1115/1.4030994
Multi-material polymer printers allow the placement of different materials within a composite. The individual material phases can be spatially arranged and shaped in an almost arbitrary fashion. Utilizing the shape memory behavior of at least one of the material phases, active composites can be 3D printed such that they deform from an initially flat plate into a curved structure. To navigate this vast design space, systematically and efficiently explorer design options, and find an optimum layout of the composite this paper presents a novel design optimization approach. The optimization approach combines a level set method for describing the material layout and a generalized formulation of the extended finite element method (XFEM) for predicting the response of the printed active composite (PAC). This combination of methods yields optimization results that can be directly printed without the need for additional post-processing steps. The proposed optimization method is studied with examples where the target shapes correspond to a plate-bending type deformation and to a localized deformation. The optimized designs are 3D printed and the XFEM predictions are compared against the experimental measurements. The design studies demonstrate the ability of the proposed optimization method to yield a crisp and highly resolved description of the optimized material layout that can be realized by 3D printing.
Toward a Unified Design Approach for Both Compliant Mechanisms and Rigid-Body Mechanisms: Module Optimization
Lin Cao, Allan T. Dolovich, Arend L. Schwab, Just L. Herder and Wenjun (Chris) ZhangJ. Mech. Des 137(12), 122301; doi: 10.1115/1.4031294
Rigid-body mechanisms (RBMs) and compliant mechanisms (CMs) are traditionally treated in significantly different ways. In this paper, we present an approach to the synthesis of both RBMs and CMs. In this approach, RBMs and CMs are generalized into mechanisms that consist of five basic modules, including Compliant Link (CL), Rigid Link (RL), Pin Joint (PJ), Compliant Joint (CJ), and Rigid Joint (RJ). The link modules and joint modules are modeled with beam and hinge elements, respectively, in a geometrically nonlinear finite element solver, and subsequently a discrete beam-hinge ground structure model is established. Based on this discrete beam-hinge model, a procedure that follows topology optimization is developed, called module optimization. Particularly, in the module optimization approach, the states (both presence or absence and sizes) of joints and links are all design variables, and one may obtain a RBM, a partially CM, or a fully CM for a given mechanical task. The proposed approach has thus successfully addressed the challenge in the type and dimensional synthesis of RBMs and CMs. Three design examples of the path generator are discussed to demonstrate the effectiveness of the proposed approach.
For the Full Article please visit ASME's Digital Collection.
Redundancy Allocation Optimization for Multistate Systems With Failure Interactions Using Semi-Markov Process
Adding redundancy is a widely used method in engineering to improve the system reliability. How to add redundancy, (i.e., to meet the reliability requirement with the minimum cost), is an interesting topic in system design. Traditionally, the optimal redundancy allocation scheme is obtained under two simplified assumptions, i.e., binary states of each component and no failure dependency between components. The binary-state assumption assumes that each component and the entire system can only have two states: fully operational and completely failed. The failure independency assumption assumes no failure interaction between components, i.e., one component failure will not affect the failure process of other components. Although those two assumptions can simplify the analysis, they may lead to inaccurate reliability predictions and thus results in doubtful and misleading redundancy allocation scheme which in fact may not meet the reliability requirement. This work proposes a method to obtain the optimal redundancy allocation scheme by using the Semi-Markov process and optimization techniques without those two simplified assumptions. The target system is a type of commonly-seen system having multiple states and failure interactions. The target system contains a main subsystem providing the required output and an auxiliary subsystem helping the main subsystem function normally, such as the rotating subsystem and the lubricating subsystem, the computer mother board and the fan, and so on. A case study of a shipboard power electronic cabinet demonstrates the applicability of the proposed approach.
For the Abstract and Full Article see ASME's Digital Collection.
Kazuko Fuchi, Philip R. Buskohl, Giorgio Bazzan, Michael F. Durstock, Gregory W. Reich, Richard A. Vaia and James J. Joo
J. Mech. Des 137(9), 091401; doi: 10.1115/1.4030876
Origami structures morph between 2D and 3D configurations, and their efficient shape reconfigurations show potential for many engineering applications. However, the enormity of the design space and the complex relationship between origami-based geometries and engineering metrics place a severe limitation on design strategies based on intuition. This work proposes a physics-based origami design method using topology optimization that determines an optimal crease pattern for a folding by adding or removing folds based on a design metric. Optimization techniques and mechanical analysis are also co-utilized to identify an action origami building block and determine the optimal network connectivity between multiple actuators.
This section includes brief descriptions of articles soon to be or recently published by the Journal of Mechanical Design. These featured articles highlight recent research developments and emerging trends in mechanical design. For Abstracts and Full Articles please see ASME's Digital Collection.