Design Synthesis Problem Research Articles

Discrete Structural Design Synthesis: A Hierarchical-Inspired Deep Reinforcement Learning Approach Considering Topological and Parametric Actions

Abstract Structural design synthesis considering discrete elements can be formulated as a sequential decision process solved using deep reinforcement learning, as shown in prior work. By modeling structural design synthesis as a Markov decision process (MDP), the states correspond to specific structural designs, the discrete actions correspond to specific design alterations, and the rewards are related to the improvement in the altered design’s performance with respect to the design objective and specified constraints. Here, the MDP action definition is extended by integrating parametric design grammars that further enable the design agent to not only alter a given structural design’s topology, but also its element parameters. In considering topological and parametric actions, both the dimensionality of the state and action space and the diversity of the action types available to the agent in each state significantly increase, making the overall MDP learning task more challenging. Hence, this paper also addresses discrete design synthesis problems with large state and action spaces by significantly extending the network architecture. Specifically, a hierarchical-inspired deep neural network architecture is developed to allow the agent to learn the type of action, topological or parametric, to apply, thus reducing the complexity of possible action choices in a given state. This extended framework is applied to the design synthesis of planar structures considering both discrete elements and cross-sectional areas, and it is observed to adeptly learn policies that synthesize high performing design solutions.

Procedure for simulating radiators of liquid cooling systems of avionics

Setting the problem. Analysis and assessment of the thermal mode of radioelectronic equipment (RES) in modern radioelectronic equipment is the most important task that is solved at the stage of design and development of the future device. Therefore, the choice of cooling system for heat-loaded structures is an integral component that must be solved in the early stages of design. The use of a modern set of computer-aided design tools (CAD) throughout the entire process of creating RES allows you to take into account all possible problem areas without resorting to the manufacture of a layout and prototype. Modeling of thermal characteristics and heat exchange processes provides for modeling procedures and ensuring the thermal regime mainly at the stage of design and topological design. At the same time, the analysis of thermal characteristics, as a rule, carried out after solving any design synthesis problem (placement, layout, selection and development of block structures, racks, etc.), if the modeling results are unsatisfactory, the design changes. Thus, in order to reduce time costs and reduce the iteration of the design process, it is necessary to solve the issues of ensuring the thermal regime in the early stages. Most often, the problem can be solved by increasing the area of the heat-emitting surface (radiator), but in some cases there is a need for additional cooling, in our case, additional use of liquid cooling. This is the reason for the relevance of the development of the methodology for modeling radiators of liquid cooling systems of the RES. Purpose. Present the methodology for simulating radiators of liquid cooling systems of the RES, the reliability of which could be confirmed by the results of analytical calculations and data of a full-scale experiment. Results. A mathematical statement has been formed, a modeling technique has been proposed and procedures for optimizing the cooling system have been implemented. Based on the developed 3D model of a radiator with a liquid cooling system, multivariable analysis procedures are implemented using modern computer-aided design tools. The liquid cooling system was optimized using modern software modeling tools, in particular the Solidworks Simulation program.

From Automated Verification to Automated Design

In this talk I will review 50 years of research on the design-synthesis problem and show how the automata-theoretic approach can be used to solve it.

Automated Multi-Level Dynamic System Topology Design Synthesis

Designing new mechatronic systems for vehicle applications is a complex and time-consuming process. The increasing computational power allows us to generate automatically novel and new mechatronic discrete-topology concepts in an efficient manner. Using state-of-the-art computational design synthesis techniques assures that the complete search space, given a finite set of system elements, is processed to find all feasible topologies. The topology generation is done by converting the design synthesis problem into a constraint satisfaction problem. Accordingly, this mathematical problem is solved by assigning the presence of components and connections to variables, whereby a set of mathematical constraints need to be satisfied. These constraints capture, in essence, formalized engineering knowledge. After solving this problem, the results are post-processed to discard redundant topologies due to isomorphism. In this paper, a newly developed software application with automated constraint generation is presented that facilitates the topology generation with multiple system levels in a loop. The scalability of the problem and the different levels of expressiveness are analyzed, and the influence of the abstraction level choice on the search space is discussed. Finally, a relevant mechatronic design study from the automotive engineering field is discussed concerning the topology synthesis of alternative electro-hydraulic actuation systems being part of new continuously variable transmission topologies, thus showing its applicability.

Finite-Element Model-Based Design Synthesis of Axial Flux PMBLDC Motors

This paper discusses the design synthesis of a permanent-magnet brushless dc (PMBLDC) machine using a finite-element (FE) model. This work differentiates itself from the past studies by following a synthesis approach, in which many designs that satisfy the performance criteria are considered instead of a unique solution. The designer can later select a design, based on comparing parameters of the designs, which are critical to the application that the motor will be used for. The presented approach makes it easier to define constraints for a design synthesis problem. A detailed description of the setting up of an FE-based design synthesis problem, starting from the definition of design variables, the FE model of the machine, how the design synthesis is carried out, and to how a design is finalized from a set of designs that satisfy the performance criteria, is included in this paper. The proposed synthesis program is demonstrated by designing a segmented axial torus PMBLDC motor for an electric two-wheeler.

Minimizing Stack and Communication Memory Usage in Real-Time Embedded Applications

In the development of real-time embedded applications, especially those on systems-on-chip, an efficient use of RAM memory is as important as the effective scheduling of the computation resources. The protection of communication and state variables accessed by concurrent tasks must provide real-time schedulability guarantees while using the least amount of memory. Several schemes, including preemption thresholds, have been developed to improve schedulability and save stack space by selectively disabling preemption. However, the design synthesis problem is still open. In this article, we target the assignment of the scheduling parameters to minimize memory usage for systems of practical interest, including designs compliant with automotive standards. We propose algorithms either proven optimal or shown to improve on randomized optimization methods like simulated annealing.

Describing the design space of mechanical computational design synthesis problems

An important challenge in mechatronic system design is to select a feasible system architecture that satisfies all requirements. This article describes (i) the necessary concepts that a system architect needs to be able to formally and declaratively describe the design space of mechanical design synthesis problems, thereby minimizing accidental complexity; (ii) how a Domain Specific Language based on the SysML modeling language and the Object Constraint Language (OCL) can be used to create this model of the design space; and (iii) an iterative process to come up with a formal model of the design space. This model describes the design space independent of any (knowledge of a) particular solving technology for the Design Space Exploration. Furthermore, the information in the model allows to select the most appropriate solving strategy for a particular design synthesis problem. The different concepts are illustrated on the example of automated synthesis of a gearbox.

A combined heuristic and indicator-based methodology for design of sustainable chemical process plants

The current emphasis on sustainable production has prompted chemical plants to minimize raw material and energy usage without compromising on economics. While computer tools are available to assist in sustainability assessment, their applications are constrained to a specific domain of the design synthesis problem. This paper outlines a design synthesis strategy that integrates two computer methodologies – ENVOP Expert and SustainPro – for simultaneous generation, analysis, evaluation, and optimization of sustainable process alternatives. ENVOP Expert diagnoses waste sources, identifies alternatives, and highlights trade-offs between environmental and economic objectives. This is complemented by SustainPro which evaluates the alternatives and screens them in-depth through indicators for profit and energy, water, and raw material usage. This results in accurate identification of the root causes, comprehensive generation of design alternatives, and effective reduction of the optimization search space. The framework is illustrated using an acetone process and a methanol and dimethyl ether production case study.

Evolutionary engineering design synthesis of on-board traffic monitoring sensors

In this paper, a formal engineering design synthesis methodology based on evolutionary computation is presented, with special emphasis on the design and optimization of distributed independent systems. A case study concerned with design of a sensory system for traffic monitoring purposes is presented, along with simulations of traffic scenarios at several levels of abstraction. It is shown how the methodology introduced is able to deal with the engineering design challenges present in the case study, and effectively synthesize novel design solutions of good quality. Moreover, when the fitness function is formulated as an aggregation of design preferences with different weights and trade-off strategies, the complete Pareto optimal frontier can be determined by the evolutionary synthesis methodology. The results of this study suggest that the approach can be useful for designers to solve challenging engineering design synthesis problems.

Robust Optimal Design of Unstable Valves

This paper is concerned with the application of robust control design concepts for the physical geometric design of electrohydraulic valves. Limitations of solenoid actuators have prevented single stage electrohydraulic valves which are simpler and more cost effective from being utilized in high flow rate and high bandwidth applications. Fluid flow force induced instability has been proposed as a means to alleviate the demand on the solenoid actuators. Previous research has demonstrated that simple changes in the valve geometry can be used to manipulate both the transient flow force as well as the steady flow force for this purpose. This paper considers the dimensional design of such ldquounstablerdquo valves to minimize the net steady flow force. The robust optimal design method, in which the design must be robust to uncertainties such as variations in operating pressure ranges and dynamic viscosity, etc., is proposed. By representing the original problem as a linear fractional transformation interconnection, the robust design problem is formulated into one of synthesizing an optimal controller for an appropriate static plant with a structured uncertainty. An algorithm for solving this design synthesis problem is proposed. A case study is conducted to compare the nominal optimal (without considering uncertainty) and the robust optimal designs. It is shown that viscosity effect is exclusively utilized in the nominal optimal design, whereas both the viscosity effect and the nonorifice flux effect are needed in the robust optimal design. The robust optimal design imposes smaller steady flow force on the spool than the nominal optimal design under perturbed situations. Based on the robust design method, an actual prototype design of the unstable valve has been developed.

Novel space-based design methodology for preliminary engineering design

This work is motivated from set-based concurrent engineering (SBCE) paradigm. In contrast to the traditional design practice, SBCE considers a broader range of design possibilities (i.e., design space) from the outset, explicitly communicates and reasons about sets of design alternatives, and gradually narrows the sets to eliminate inferior alternatives until a final solution remains. Thus, the key to success of SBCE is the proper implementation of the space representation method to define the possible design region, the space mapping method to obtain the performance space achievable by the initial design space, and the space narrowing method to eliminate infeasible subspaces from the initial design space. This paper proposes a novel space-based design methodology for preliminary engineering design. The main characteristic features of our approach are to incorporate the designer’s preference structure with degrees of desirability in specifying both design space and performance space, and to find a ranged set of design solutions that satisfy changing sets of performance requirements through set-to-set space mapping from design space to performance space and space narrowing to eliminate infeasible design subspaces. Central to the proposed design methodology is the integration of meta-modeling techniques, modified fuzzy arithmetic, design of experiment (DoE), robust design techniques, and uncertainty analysis. A preliminary parametric design example of a vehicle side impact beam is demonstrated to show the effectiveness of the proposed methodology and its availability in the large-scale multi-objective design synthesis problem.

A Configuration Synthesis Method of Mechanism Systems Using Motion Marker Sequence and Planar Cells

We discuss the design synthesis problem of mechanical systems with an example of configuration design of chain-type mechanisms. Mechanisms are used as system elements in various products, and each mechanism system is composed of mechanism pairs. In the design process, the functional structure of mechanisms and their spatial configuration are determined. The former means the sequential chains of pairs, and the latter means the arrangement and geometry of mechanism elements. We propose a framework for simultaneously considering both issues for designing a certain level of complicated mechanism systems. The concepts of marker sequences and cell-based planar representation are introduced to manipulate a design. Each marker corresponds to an abstracted representation of a particular mechanism element. Each cell corresponds to a region for such a marker. The computational design process is organized by generating marker sequences, assigning markers to cells, refining both of them, and embodying the configuration and arrangement of mechanisms. Finally, a prototype CAD system implemented under the proposed concepts is applied to several design examples. In the conclusion, we show the necessity of fictitious media for representing and computerizing the conceptual and configurational phases of a design process.

A new synthesis approach for feedback neural networks based on the perceptron training algorithm

In this paper, a new synthesis approach is developed for associative memories based on the perceptron training algorithm. The design (synthesis) problem of feedback neural networks for associative memories is formulated as a set of linear inequalities such that the use of perceptron training is evident. The perceptron training in the synthesis algorithms is guaranteed to converge for the design of neural networks without any constraints on the connection matrix. For neural networks with constraints on the diagonal elements of the connection matrix, results concerning the properties of such networks and concerning the existence of such a network design are established. For neural networks with sparsity and/or symmetry constraints on the connection matrix, design algorithms are presented. Applications of the present synthesis approach to the design of associative memories realized by means of other feedback neural network models are studied. To demonstrate the applicability of the present results and to compare the present synthesis approach with existing design methods, specific examples are considered.

Global optimization using accurate approximations in design synthesis

The design variable space of a design synthesis problem may contain multiple local optima. In the approximation concepts approach to design synthesis, the design objective and constraint functions are approximated in order to reduce the overall cost. If the approximations accurately capture the actural behavior of the objective function and constraints, then the approximate design variable space may also contain local optima. In this work, a multistart optimization algorithm is used to search for the global optimum of the actual design using just a few design cycles. Example problems are presented to illustrate the methodology set forth.

On the complexity of the design synthesis problem

Presents and analyzes a formal model of a design process, emphasizing the synthesis part. The design artifact description is identified as an algebraic structure. The desired function and constraints are mapped to the artifact description using an evolutionary process that can be visualized as a feedback loop of analysis, synthesis and evaluation. A special case of the synthesis activity, called the basic synthesis problem (BSP), is addressed. The BSP is shown to be NP-complete. As a consequence, tractability can be obtained by enforcing constraints on the artifact structure. As such, the authors present a model having an element of descriptive design theory that is also a framework for the future development of computational support systems and automatic design tools.

Experiments in process synthesis via mixed-integer programming

Mixed-integer programming is the natural formulation of many process design synthesis problems. Up to now mixed-integer programming algorithms have been applied successfully for solving simplified synthesis problems involving Boolean variables, with the assumption of separability and linearity of constraints and/or objective function. This work aims at going still further in the solution of mixed problems with non-Boolean variables, nonlinear and nonseparable constraints and criteria. A general-purpose algorithm is presented and illustrated by two examples of processes. The synthesis of a heat exchanger network shows the capability of the algorithm to solve this problem via a simultaneous optimization of the structure and temperatures, without assuming a minimum temperature difference. The number of stages and the operating conditions of a multistage reactor for esterification of fatty acid are then optimized, under a purity constraint of the ester, computed by simulation. This second example shows that it is possible to handle nonlinear and implicit constraints.

Locating Point Synthesis in Fixture Design

A method for selecting and positioning locating points in fixture design is proposed. The types of fixture applications considered are for prismatic workpieces that use the 3-2-1 locating principle as the general structure of the fixture. The design synthesis problem uses six rules based on machining practice to select the secondary and tertiary locating datums for a workpiece stored in a boundary representation solid modeler. Once these edges are selected, kinematic rules are used to determine the position of the locating points so as to allow accessibility when loading and unloading a workpiece. Examples are given to illustrate this method, as well as the positioning of support points on the primary datum plane.