Automated Design Synthesis Research Articles

Additive manufactured continuum mechanisms based on shape-programmable and micro-sized building blocks

ABSTRACT Micro-additive manufacturing techniques have the potential to meet the demand for miniaturised functional components for minimally invasive surgical instruments. These techniques create monolithic, compliant mechanisms with micro-sized free-form structures that can be tailored to patient-specific surgical procedures. The automated design synthesis of the mechanisms using building blocks results in structures that are shape-programmable. This is achieved through an algorithmic-based computational workflow, which automatically converts user-specified 2D and 3D curves into discrete curve segments. The actuated motion of the mechanisms can be designed to move in a specific way, both forwardly and inversely. The mechanisms are manufactured using micro-laser powder bed fusion and hardenable stainless steel 17-4 PH. By carefully selecting the process parameters, it is possible to 3D-print micro-sized features such as a compliant beam thickness of 80 μm and an actuation hole of 100 μm. Both 2D planar curved mechanisms and 3D spatial curved mechanisms have been implemented and experimentally validated.

Deep Generative Models in Engineering Design: A Review

AbstractAutomated design synthesis has the potential to revolutionize the modern engineering design process and improve access to highly optimized and customized products across countless industries. Successfully adapting generative machine learning to design engineering may enable such automated design synthesis and is a research subject of great importance. We present a review and analysis of deep generative machine learning models in engineering design. Deep generative models (DGMs) typically leverage deep networks to learn from an input dataset and synthesize new designs. Recently, DGMs such as feedforward neural networks (NNs), generative adversarial networks (GANs), variational autoencoders (VAEs), and certain deep reinforcement learning (DRL) frameworks have shown promising results in design applications like structural optimization, materials design, and shape synthesis. The prevalence of DGMs in engineering design has skyrocketed since 2016. Anticipating the continued growth, we conduct a review of recent advances to benefit researchers interested in DGMs for design. We structure our review as an exposition of the algorithms, datasets, representation methods, and applications commonly used in the current literature. In particular, we discuss key works that have introduced new techniques and methods in DGMs, successfully applied DGMs to a design-related domain, or directly supported the development of DGMs through datasets or auxiliary methods. We further identify key challenges and limitations currently seen in DGMs across design fields, such as design creativity, handling constraints and objectives, and modeling both form and functional performance simultaneously. In our discussion, we identify possible solution pathways as key areas on which to target the future work.

RANGE-GAN: Design Synthesis Under Constraints Using Conditional Generative Adversarial Networks

Abstract Typical engineering design tasks require the effort to modify designs iteratively until they meet certain constraints, i.e., performance or attribute requirements. Past work has proposed ways to solve the inverse design problem, where desired designs are directly generated from specified requirements, thus avoid the trial and error process. Among those approaches, the conditional deep generative model shows great potential since 1) it works for complex high-dimensional designs and 2) it can generate multiple alternative designs given any condition. In this work, we propose a conditional deep generative model, Range-GAN, to achieve automatic design synthesis subject to range constraints. The proposed model addresses the sparse conditioning issue in data-driven inverse design problems by introducing a label-aware self-augmentation approach. We also propose a new uniformity loss to ensure generated designs evenly cover the given requirement range. Through a real-world example of constrained 3D shape generation, we show that the label-aware self-augmentation leads to an average improvement of14% on the constraint satisfaction for generated 3D shapes, and the uniformity loss leads to a 125% average increase on the uniformity of generated shapes' attributes. This work laid the foundation for data-driven inverse design problems where we consider range constraints and there are sparse regions in the condition space.

Large-Scale Three-Dimensional Anisotropic Topology Optimization of Variable-Axial Lightweight Composite Structures

Abstract Variable-axial fiber-reinforced composites allow for local customization of fiber orientation and thicknesses. Despite their significant potential for performance improvement over the conventional multiaxial composites and metals, they pose challenges in design optimization due to the vastly increased design freedom in material orientations. This paper presents an anisotropic topology optimization method for designing large-scale, 3D variable-axial lightweight composite structures subject to multiple load cases. The computational challenges associated with large-scale 3D anisotropic topology optimization with extremely low volume fraction are addressed by a tensor-based representation of 3D orientation that would avoid the 2π periodicity of angular representations such as Euler angles, and an adaptive meshing scheme, which, in conjunction with PDE regularization of the density variables, refines the mesh where structural members appear and coarsens where there is void. The proposed method is applied to designing a heavy-duty drone frame subject to complex multi-loading conditions. Finally, the manufacturability gaps between the optimized design and the fabrication-ready design for tailored fiber placement (TFP) is discussed, which motivates future work toward a fully automated design synthesis.

PaDGAN: Learning to Generate High-Quality Novel Designs

Abstract Deep generative models are proven to be a useful tool for automatic design synthesis and design space exploration. When applied in engineering design, existing generative models face three challenges: (1) generated designs lack diversity and do not cover all areas of the design space, (2) it is difficult to explicitly improve the overall performance or quality of generated designs, and (3) existing models generally do not generate novel designs, outside the domain of the training data. In this article, we simultaneously address these challenges by proposing a new determinantal point process-based loss function for probabilistic modeling of diversity and quality. With this new loss function, we develop a variant of the generative adversarial network, named “performance augmented diverse generative adversarial network” (PaDGAN), which can generate novel high-quality designs with good coverage of the design space. By using three synthetic examples and one real-world airfoil design example, we demonstrate that PaDGAN can generate diverse and high-quality designs. In comparison to a vanilla generative adversarial network, on average, it generates samples with a 28% higher mean quality score with larger diversity and without the mode collapse issue. Unlike typical generative models that usually generate new designs by interpolating within the boundary of training data, we show that PaDGAN expands the design space boundary outside the training data towards high-quality regions. The proposed method is broadly applicable to many tasks including design space exploration, design optimization, and creative solution recommendation.

Automated Synthesis of Passive Dynamic Brachiating Robots Using a Simulation-Driven Graph Grammar Method

Passive dynamic systems have the advantage over conventional robotic systems that they do not require actuators and control. Brachiating, in particular, involves the swinging motion of an animal from one branch to the next. Such systems are usually designed manually by human designers and often are bio-inspired. However, a computational design approach has the capability to search vast design spaces and find solutions that go beyond those possible by manual design. This paper addresses the automated design of passive dynamic systems by introducing a graph grammar-based method that integrates dynamic simulation to evaluate and evolve configurations. In particular, the method is shown to find different, new solutions to the problem of the design of two-dimensional passive, dynamic, continuous contact, brachiating robots. The presented graph grammar rules preserve symmetry among robot topologies. A separation of parametric multi-objective optimization and topologic synthesis is proposed, considering four objectives: number of successful swings, deviation from cyclic motion, required space, and number of bodies. The results show that multiple solutions with varying complexity are found that trade-off cyclic motion and the space required. Compared to research on automated design synthesis of actuated and controlled robotic systems, this paper contributes a new method for passive dynamic systems that integrates dynamic simulation.

Comparing Strategies for Topologic and Parametric Rule Application in Automated Computational Design Synthesis1

Graph grammars are used for computational design synthesis (CDS) in which engineering knowledge is formalized using graphs to represent designs and rules that describe their transformation. Most engineering tasks require both topologic and parametric rules to generate designs. The research presented in this paper compares different strategies for rule application to combine topologic and parametric rules during automated design synthesis driven by a search process. The presented strategies are compared considering quantity and quality of the generated designs. The effect of the strategies, the selected search algorithm, and the initial design, from which the synthesis is started, are analyzed for two case studies: gearbox synthesis and bicycle frame synthesis. Results show that the effect of the strategy is dependent on the design task. Recommendations are given on which strategies to use for which design task.

An evolutionary geometric primitive for automatic design synthesis of functional shapes: The case of airfoils

A novel self-adaptive geometric primitive for functional geometric shape synthesis is presented. This novel geometric primitive, for CAD use, is specifically designed to reproduce geometric shapes with functional requirements, such as the aerodynamic and hydrodynamic ones, once the functional parameters are furnished. It produces a typical CAD representation of a functional profile: a set of Bezier curves. The proposed primitive follows a generate-and-test approach and takes advantage of the use of a properly designed artificial neural network (BNN). It combines the properties of a geometric primitive and the capability to manage the engineering knowledge in a specific field of application. The proposed evolutionary primitive is applied to a real engineering application: the automatic synthesis of airfoils. Some examples are simulated in order to test the effectiveness of the proposed method. The results obtained by an original prototypal software are presented and critically discussed.

An automated design synthesis method for multi-trailer articulated heavy vehicles

This paper presents an automated design synthesis method for multi-trailer articulated heavy vehicles (MTAHVs) to coordinate the trade-off between the manoeuvrability and stability. Conventionally, the design synthesis of MTAHVs is based on trial and error approaches. This is difficult, time-consuming and tedious. To tackle this problem, a design method is proposed. The method has the following features: 1) vehicle modelling, performance evaluation, and design selection are implemented by computers; 2) a MTAHV model is introduced in the design optimisation; 3) in the closed-loop simulation, testing manoeuvres are emulated such that a driver model 'drives' the virtual MTAHV to follow a predefined trajectory. To test the method, the combination of a tractor and two semitrailers are optimised. It is indicated that the conflicting criteria of manoeuvrability and stability can be simultaneously improved. The approach may identify desired design variables and predict performance envelopes in early design stages of MTAHVs.

Shape optimization of a MHD generator based on pressure drop and power output constraints

The optimum shape and axial variation of electrical loading (power ratio) of a MHD generator are obtained for a prescribed fractional power output and pressure drop along the generator. These optimum solutions are obtained using the ADS (Automatic Design Synthesis) optimization code. The objective function (duct size) for the prescribed design conditions was obtained by Carter for a quasi 1-D flow with a simple power law model for electrical conductivity. In this work, a generalized numerical optimization scheme, capable of incorporating any arbitrary electrical conductivity model is presented. The numerical solutions obtained using this scheme is validated with analytical results obtained by Carter. We demonstrate the use of this scheme to minimize the volume of a duct with a hypothetical power law model incorporating the effect of residence time (function of inlet Mach number) on the electrical conductivity. We further demonstrate the use of this scheme for obtaining the optimum shape of a MHD generator using an E-beam generated ionized gas stream.

Optimization of orthotropic rectangular laminates with weak interfaces under buckling

Abstract The buckling of an orthotropic laminate with weak interfaces is investigated. The state-space formulations established directly from the three-dimensional theory of elasticity are employed. The present analysis is effective because the final scale of the resulting equations always keeps the same in spite of the number of layers as well as the number of weak interfaces in the laminate. To maximize the critical buckling stress of laminates with weak interface, the optimal design are performed to find the optimal design of microstructural parameters using ADS (automated design synthesis).

Design synthesis from interaction and state-based specifications

Interaction-based and state-based modeling are two complementary approaches of behavior modeling. The former focuses on global interactions between system components. The latter concentrates on the internal states of individual components. Both approaches have been proven useful in practice. One challenging and important research objective is to combine the modeling power of both effectively and then use the combination as the basis for automatic design synthesis. We present a combination of interaction-based and state-based modeling, namely, live sequence charts and Z, for system specification. We then propose a way of generating distributed design from the combinations. Our approach handles systems with intensive interactive behaviors as well as complex state structures

Function-based design synthesis approach to design reuse

In this paper, a design reuse framework with a function-based design synthesis approach is proposed in the context of conceptual product development. Previous researches in design reuse have lacked a comprehensive functional base for knowledge representation and reasoning. The method presented in this paper uses a function-based product information model and a multiple objective optimization model to achieve design reuse. The information model is dependent on a functional core which is the key element vector. It is capable of modeling product information with sufficient abstraction, which in turn facilitates intelligent construction of product platforms. The multiple objective optimization method carries out automated design synthesis and evaluation subject to various design constraints. The approach has been applied in the design of the fan filter unit, a key clean room device. It has achieved intelligent design reuse in product conceptual design with significant rapidity and solution variety.

Critical thrust force at delamination propagation during drilling of angle-ply laminates

A new formulation for the critical thrust force necessary to propagate the delamination generated during the drilling operation of an antisymmetric angle-ply laminate is proposed by modeling the delamination zone as an elliptical shape. The critical thrust force is analytically derived with the consideration of bending, twisting and mid-plane extension of the delamination zone. And then to maximize the critical thrust force, the optimal design of an angle-ply laminate is performed to find the optimal number of fiber per millimeter, optimum diameter of fiber and optimum lamination angle using ADS (Automated Design Synthesis).

Critical thrust force at delamination propagation during drilling of angle-ply laminates

A new formulation for the critical thrust force necessary to propagate the delamination generated during the drilling operation of an antisymmetric angle-ply laminate is proposed by modeling the delamination zone as an elliptical shape. The critical thrust force is analytically derived with the consideration of bending, twisting and mid-plane extension of the delamination zone. And then to maximize the critical thrust force, the optimal design of an angle-ply laminate is performed to find the optimal number of fiber per millimeter, optimum diameter of fiber and optimum lamination angle using ADS (Automated Design Synthesis).

유한차분법과 ADS(Automated Design Synthesis)를 이용한 광픽업 액추에이터의 동특성 개선

Recently, there have been many researches and developments in optical disc drive by high density of track and high access speed. Therefore, the optical pick-up should guarantee highly accurate dynamic characteristics for the wide bandwidth in order to cope with this trend. These demands for optical pick-up actuator can be solved by improvements of lens-holder through the following methods. The first way is the analysis of the sensitivity matrix of design variables for vibration modes after appropriate design parameters are selected like shapes and local dimensions of a lens-holder. The second way is the optimization of design variables by calculating sequential linear programming after the problem of extending bandwidth are converted to problem of minimizing adequate objective function. In the result, modified FE model is obtained through several iterations by finite difference scheme(FDS). While results of the first way show better convergence of the target frequency, the second result shows better reduction of mass increase.<br/>

The A-Design approach to managing automated design synthesis

This paper presents an approach to managing the complexities of design synthesis techniques. Often computational approaches model design as a search process or as an optimization problem. The nature of such techniques, however, does not allow the user to interact with the search process once it has begun. Furthermore, traditional computational search lacks the ability to learn from experience. While computational search techniques have the ability to search many design alternatives quickly, the human engineer can often arrive at a more elegant and robust solution by applying heuristics learned from past experiences. The method introduced here improves the capabilities of design synthesis methods by allowing for user input and by making decisions based on previous experience. In studying past candidate designs the process learns to be more effective in searching for solutions. Results show how such a technique improves the quality of designs and efficiency of an existing search process.

Shape optimization, model updating and empirical modeling applied to the design synthesis of a heavy truck side guard

This work describes the use of different optimization approaches aiming at the automatic design synthesis of a heavy truck side guard (bumper). At a preliminary stage, an existing version of the side guard undergoes a geometrical simplification in order to improve manufacturability and assemblability aspects of the design. This geometrical simplification is achieved through a fully automatic shape optimization procedure, taking into account that key mechanical features of the original configuration are to be kept in the simplified design. In the sequence, the numerical model of the simplified version of the bumper is updated to match the experimental results obtained through modal testing of a prototype in laboratory conditions. For comparison purposes, two automatic optimization techniques are used in the modal matching procedure. After achieving the desired configuration changes (by means of shape optimization) and a reliable model representing the system's behaviour (by means of the modal matching procedure), the design of the heavy truck side bumper is improved, again by using an automatic approach based on numerical optimization techniques. The purpose of this optimization procedure is to tune the natural vibrating frequencies of the side bumper so that its behaviour with respect to noise, vibration and harshness reaches predefined performance levels. Aiming at future design improvement, to be held in a multidisciplinary context that accounts for crashworthiness issues, empirical models (non-linear response surfaces) are developed for the implementation of such optimization procedure. Finally, the general conclusions and perspectives for future research work are presented, stressing at the usefulness of numerical optimization techniques in supporting the automation of an entire design procedure, from conceptual definition to multidisciplinary performance improvement.

Geometric event constraints: design synthesis by a multibody Lagragian type I formalism

This paper introduces the geometric event constraint for the automatic design synthesis of mechanisms. Using a Lagrange type I formalism event constraints and dummy particles can efficiently find the optimal joint configuration in mechanism design studies. Two examples are presented using nonholonomic and holonomic event constraints. The holonomic system is simulated for a locus of optimal performance joint locations during mechanism articulation, providing both synthesis and analysis in one simulation. Desired system response, dependent on joint configurations, is obtained using standard multibody dynamics software avoiding additional special purpose code or optimisation routines.

A-Design: An Agent-Based Approach to Conceptual Design in a Dynamic Environment

This paper provides an introduction to a new design methodology known as A-Design, which combines aspects of multi-objective optimization, multi-agent systems, and automated design synthesis. The A-Design theory is founded on the notion that engineering design occurs in interaction with an ever-changing environment, and therefore computer tools developed to aid in the design process should be adaptive to these changes. In this paper, A-Design is introduced along with some simple test problems to demonstrate the capabilities of different aspects of the theory. The theory of A-Design is then shown as the basis for a design tool that adaptively creates electro-mechanical configuration designs for changing user preferences.