This paper proposes a method for determining optimal back-pressure profile in forging using a sequential approximate optimization (SAO). We consider that back-pressure profile that varied through the stroke is valid. In order to determine the optimal back-pressure profile, this object is formulated as a multi-objective optimization. Thus, the unfilled area is taken as the first objective function, and the forming energy is also taken as the second one. Computer-aided engineering (CAE) is used, but the numerical simulation in the forging is so expensive that the SAO using the radial basis function network is adopted. The pareto-frontier is identified with a small number of simulation runs. In this paper, the axial symmetric forward extrusion of aluminum alloy is used as the forming model. First, the experiment is carried out to construct the finite element analysis model. Through the numerical simulation and the experiment, we can confirm that the product quality is improved with the optimal back-pressure profile.
The self-organizing map (SOM) is a kind of artificial neural networks that forms a topological map to cluster data as well as to reduce dimensionality of the data based on unsupervised learning. In this study, the SOM is introduced to the quality inspection process of beverage cans where the input vector with large dimension is composed of a series of frequency response data that is obtained by the magnetic hammering test. In response to the input vectors, the system predicts the internal pressure of cans in a way that a certain unit comprising the learned network responds to an input vector having minimum distance with the weight vector of the selected unit. Test results have shown that the SOM can roughly identify the difference of the internal pressure of cans, however, the estimate accuracy is not satisfactory in comparison to the prediction results obtained by the system which is based on the k-nearest neighbor algorithm. Toward the further improvement of the suggested classification system based on SOM, the investigation into the well-suited choice of input vectors, proper arrangement of the network parameters are necessary.
The performance of machine learning such as support vector machines and radial basis function neural networks depends on parameters, for example the width of Gaussian function and a regularization parameter. One of most popular methods for estimating parameters is cross validation (CV) test, however CV test is usually time consuming. In this research, we propose an effective learning method using bagging and boosting in order to reduce the burden on choosing the width of Gaussian function as well as the sensitivity to it. Additionally, we show that comparing with the calculation time by a single machine, the one by the proposed learning method can be improved without losing a performance of generalization ability through several numerical experiments.
In this paper, we will examine the possibility of availability of predictive control for complicated modeling by apply experimental learning algorithm using RBF Network to model predictive control. As an experiment device for examining, we make model in imitation of a crane and use this.
A turbocharger for gasoline engines is exposed to low-cycle thermal field. It is important to prevent a turbocharger from low-cycle thermal fatigue fracture. In conventional design methods, engineers design it using their knowledge and doing trial and error procedures. However, it is difficult to design the optimal shape by trial and errors. In addition, it requires large amount of trials since a shape of turbocharger is complicated. In the present paper, we develop a new design frame work to reduce low-cycle thermal stress using a non-parametric shape optimization method that can automatically and freely modify surfaces. We then apply our new design frame work to an exhaust manifold of turbine housing integrated exhaust manifold.
This paper deals with a numerical analysis method based on a time integration technique and the H1 gradient method for solving nonparametric boundary shape optimization problems of domain boundaries in which boundary value problems of partial differential equations are defined. The H1 gradient method has been developed by applying the gradient method in a Hilbert space. In the H1 gradient method, the domain variation that minimizes the objective functional is obtained as a solution to a boundary value problem of a linear elastic continuum defined in the design domain loaded with traction in proportion to the negative shape gradients on the design boundary. The optimized shape is obtained as the solution to first-order ordinary differential equations with the solution of the H1 gradient method. The effectiveness of the proposed method is demonstrated through numerical examples in a heat conduction problem and a structural problem.
The present paper describes an evaluation method of the shape derivatives, which is defined as the Frechet derivative with respect to the domain variation, of cost functions in a shape optimization problem of a domain in which a boundary value problem of partial differential equation is defined. Instead of the known method using the equation of boundary integral type, an equation of domain integral type is proposed to evaluate the shape derivative. This method is applied to the mean compliance minimization problem of linear elastic body. Numerical example shows the validity of the method.
Firefly algorithm (FA) has been used as analysis procedure that can obtain not only a global optimal solution but also local optimal solutions by setting multiple computational parameters. Here, to simplify the setting of these computational parameters and improve the search performance of the high-dimensional problem, we implement two operations that a dimensionless distance of design variable space, and the clustering of design variable space. In this study, FA including these operations is applied shape optimization problem of continuum shell structure, and then the effectiveness and stability of the solution using FA are indicated through these results.
In this study, we deal with multi-objective combinatorial optimization problems because many real problems with discrete structures can be formulated for combinatorial optimization problems and there are plural objective functions in real problems. We focus on metaheuristics as optimization method for combinatorial optimization problems. In metaheuristics, how to use the search history information and the interactions among search points at making the neighborhoods and moving search points are important. We proposed a multi-point search method which is based on Tabu Search at making the neighborhoods and introducing new interactions among search points at moving search points. In this study, we analyze the method qualitatively and propose a new moving strategy based on diversification and intensification. We examine the method with the new moving strategy and validate utility of the strategy through the numerical experiment.
Particle Swarm Optimization (PSO) is one of the evolutionary computations. Position vectors of particles in a swarm represent candidate solutions of the optimization problem. The position vectors, in the original PSO, are updated with the position vectors of the global and personal best particles. The global and personal best particles denote the particle that all particles and each particle have found ever during search process, respectively. In this study, particle position vectors are updated with the second global or personal best particles, in addition to the position vectors of the global and personal best particles. The present PSO algorithms are applied for the truss structure optimization problem. The results reveal that the present algorithm can find better optimal solution than that of the original PSO.
This paper focuses on distributed LED lighting systems composed of multiple LED lights. The lighting patterns (ON/OFF mode) are determined such that a supplying illuminance distribution is similar to a desired one. In order to determine the lighting patterns, we utilize a distributed optimization algorithm. Then, we develop an experimental system with 25 LED units, and verify the effectiveness of our algorithm through several experiments.
When we measure the coordinate position exactly in the three-dimensional space from a target threshold, three range-finding devices are required. That is non-efficiency on a cost side. We suggest using Levenberg-Marquardt method to reduce the number of the devices to be necessary from three to two for this problem. In this study, we use Kinect for Windows version 1 provided by Microsoft Corporation for as a range-finding device. To estimate coordinate we use coordinates of two Kinect devices and each distance to object. We also review initial coordinate of the objects.
Recently, there are a lot of reports for evaluation of landscape by using fractal analysis. Although color information is quite important on design evaluation, they do not deal with it. In this paper, a new strategy for fractal analysis that can deal with color information is proposed. We also have applied it to design evaluation for product and use it as a tool for Kansei Design.
'Tiling' is one of the classic themes of geometry, which is also referred to as 'tessellation'. In this study, we deal with an engineering optimization problem of tiling, that is, the problem of filling an arbitrary-shaped region with tiles. In contrast to the mathematical tiling or tessellation, the joint part between tiles and the cutting of tiles for shaping are taken into account. A general formulation of the problem and its specific case of a practical problem are developed. The approach discussed in this article is not elegant from the theoretical or mathematical viewpoint; however, the obtained optimal tiling examples well demonstrate the significance of this type of primitive engineering optimization approach.
The aim of this study was to optimize grasped object shape to improve its grasping ease with considering the interaction of multiple shape parameters. The shape parameters used in this study were the upper diameter d_U, lower diameter d_L, ellipticity e, and inclined angle of central axis θ. First, two levels factorial experiment was performed with five male subjects. Multi-way ANOVA was conducted for the measured subjective scores, and the result showed that the main effects of the upper diameter, ellipticity, and inclined angle are statistically significant, and the significance probability of the lower diameter was relatively low. In addition, the interaction of upper diameter and inclined angle and that of ellipticity and inclined angle had significant effects. The four parameters were assigned to the L_<27> orthogonal array, and the subjective scores were measured with six male and six female subjects. The response surface of the grasping ease was approximated, and the optimum shape was determined.
Ankle-foot orthosis (AFO) is a device widely prescribed for hemiplegic patients to assist their walking. Conventionally, AFOs are expected to have sufficiently high ankle joint stiffness to prevent involuntary plantar flexion during the swing phase. Such a high joint stiffness, however, disturbs the natural dorsal flexion due to the forward tilting of lower leg during the stance phase. The authors have proposed a concept of a variable stiffness AFO that is able to alter the ankle stiffness of AFO in accordance with the change of walking phase. This report discusses optimal shape design problems of three-dimensional monolithic AFO structure combining a stiffness switching mechanism with a joint structure of shoehorn AFO.
The visualization of Kansei value as requirements and the realization of its value for making a product idea in a conceptual design are desired. Therefore, A. Sato and H. Hasegawa have proposed the Idea Creation Support System (ICSS) - for an idea creation with Kando understanding process through WOM (Word Of Mouth) effectiveness. However, ICSS could not realize a visualization of a requirement according to topology, i.e., shape and layout, through understanding Kansei value clearly. For realization of this problem, we propose the topology creation method which consisted of SIMP method as a topology optimization and the Kansei information understanding of a topology through Electroencephalogram (EEG), and also discuss the system configuration of ICSS with the topology creation method.
Flow diverter stent (FD) is one of an efficient medical device or intracranial aneurysm. FD is a cylindrical device woven from metallic wires. It blocks the inflow entering an aneurysm to make blood coagulation in the aneurysm. But the device has a risk of blood coagulation as well in the parent artery attributable to chemical reactions between metal and blood. Therefore, we have developed the optimization program for mesh design of FD with simulated annealing and computational fluid dynamics to improve the treatment effect maintaining low wire density. In this report, we investigate the relationship between hemodynamic parameters and mesh design to determine the objective function for FD optimization.
Discus throwers develop and adjust their skills depending on the discus that is being used. The optimal skill is applicable to the discuses used in their training and those used in competitions. However, their performance could be maximized if their skill and their equipment were optimized simultaneously. Here, the objective function for optimization is the flight distance. Longer flight distance is better. Fourteen design variables are considered. Eight of the fourteen are concerned with the skill of the thrower. They determine the launch conditions, which are controlled by the thrower when he or she throws. The other six variables are concerned with the design of the equipment. The optimization is carried out with the aid of a genetic algorithm. It was found that the flight distance obtained by concurrent optimization of both skill and equipment is 79.1 meters, while the skill-optimized flight distance is 76.9 meters.
Lotus leaves are well known for its high water repellency. Lotus leaves have microstructure composed of microrods on a surface and when a water drop sit on a leaf, microrods keep air layer between surface and drop. So, drop contacts by extremely low area and it can be seen as if it floats on a surface. The microrods are not uniformed; does not have same diameter, height. In this paper, focusing on this characteristic and consider of its effect on water repellency by observation and simulations.
We formulate an optimization method for flow field based on transient information of the flow field. The flow field is solved by the lattice Boltzmann method. We use a level-set function in order to represent the boundary clearly, and apply non-slip boundary conditions at the boundary. We formulate the sensitivity considering that the boundary moves during the process of optimization. We also show a numerical examples.
We propose a new method to study the effectiveness of hemodynamic factors to determine artery shape. At first, multi-objective optimization that set five kinds of determinant factors as objectives was performed to obtain optimized shapes on the basis of seven actual carotid artery bifurcation shapes using computational fluid dynamics and genetic algorithms. Those factors are a) maximum Wall Shear Stress (WSS), b) minimum WSS, c) WSS gradient, d) WSS temporary gradient and e) integration of artery radius. Then, the optimized shapes were checked if they include the original one or not. After all optimizations, the most matched factors were selected. As a result, in the case of minimizing both maximum WSS and integration of artery radius, all 7 original shapes were included in the optimized ones. This combination was the best in six tested combinations. It became clear this combination plays the most important role than the other tested combinations. We confirmed our method is very useful to examine effectiveness of unknown biological adaptation.
Efficient mixing or kneading process of materials having different material properties, e.g., density, viscosity, are required in many industries. In general mixing process contains many parameters to be adjusted for homogeneous and quick material mixing and much time and cost are needed to determine adequate parameters. So mixing process is considered as one of suitable manufacturing process to apply numerical optimization technique. Recently unique computational fluid dynamics software using Moving Particle Simulation (MPS) method have been developed. MPS is a kind of particle method and have been developed as an incompressible fluid flow solver. MPS is completely Lagrangian formulation and doesn't need spatial mesh definition. So MPS has many advantages especially for mixing simulation. In this paper efficient optimization technique using commercial MPS software Particleworks^[○!R] is demonstrated.
In this article, we performed the examination by the finite element method about optimization in the space-time temperature distribution of the unsteady two-dimensional heat transfer system. We used package software COMSOL Multiphysics. In order to apply Nelder-Mead to unsteady problem, the two-dimensional heat transfer equation was transformed to steady three-dimensional heat transfer equation.
This paper presents a truss topology optimization method for finding the minimum compliance design in which only a limited number of different cross-section sizes are used. The member cross-sectional areas are considered either discrete design variables or continuous design variables. In both cases the optimization problem of interest can be reduced to a mixed-integer second-order cone programming problem. The global optimal solution is then computed by using an existing algorithm based on a branch-and-cut method.
A layerwise optimization (LO) approach was previously proposed by the author for the lay-up optimization problem of laminated composites, and has been extended in many applications. Various structural analysis methods, including semi-analytical Ritz method, self-made FEM and commercial FEM, are accommodated in the LO optimization. In such lay-up design problems, they usually cause rapid increase in computation time due to the search for optimum solutions of the multi-dimensional space, when design variables are taken to be the fiber orientation angles directly in all K layers. The LO makes it possible, however, that this multi-dimensional optimization problem can be reduced into only K times repetition of one-dimensional search. The theoretical background, however, has not been discussed enough despite wide applicability and accurate approximation. The present report takes up this basic issue and attempts to clarify the reason why an algorithm of the LO approach works effectively.
Aggregative gradient-based multiobjective optimization (AGMO) is a methodology for obtaining Pareto optimal solutions based on local search techniques. AGMO typically uses the weighing-sum technique, with weighting coefficients adaptively given during the optimization process. An AGMO method can handle large-scale problems that include numerous constraints and design variables, which are problematic when using meta-heuristic techniques. This paper proposes techniques to enhance the diversity of search points during the optimization process in an AGMO method, by introducing distance constraints and adaptive adjustment of the number of search points, so that well-distributed Pareto solutions are obtained. The proposed method is applied to example problems to illustrate its effectiveness.
In a head positioning control system of hard disk drives, multi-stage notch filters are used in order to stabilize mechanical resonant modes of the controlled system. Conventionally, the parameters of these filters are tuned by try and error method in advance of the mass production of HDDs. By this tuning method, it is difficult to prevent the occurrence of defective products caused by the variation of mechanical resonant modes. In order to overcome this problem, we have developed an auto-tuning method of these parameters in production line by multi-objective optimization.
The present paper proposes a multi-objective optimization technique for smart laminated composites to maximize two conflicting objectives. The first objective is the performance of active vibration control of smart composite with piezoelectric (PZT) actuators. The second is the fundamental frequency of smart structures related to the performance of passive vibration control. Both performances of active and passive vibration control are maximized simultaneously. The vibration suppression of smart structures strongly depends on both actuator placements and vibration mode shapes. It is possible to design vibration mode shapes for laminated fibrous composites since their anisotropy for whole thickness is tailorable by arranging fiber orientation angle in each layer. This allows the smart structure with laminated composite to archive higher performance of vibration suppression than those with isotropic materials. However, the optimized structure results in lower natural frequencies than composites with typical fiber orientation angles since an effective input of control force from actuators is realized for the structure with lower stiffness. This reveals that there is a trade-off relation for smart composite structures between the performance of active vibration suppression and natural frequencies. To disclose this relation, the present study applies the effective multi-objective optimization technique, the refined non-dominated genetic algorithm (NSGAII), and obtains Pareto optimal solutions. Calculated results are successfully validated by a comparison with those from the real-tune control experiment where a laser excitation technique which is effective to small sized structures is used.
The purpose of this study is to generate the gait of a two-legged robot to avoid obstacles. It is expected that two-legged robot can avoid obstacle more smoothly in the same way that animal and human adjust stride naturally to step over obstacles. Stepping points are determined optimally for walking. The gait generation problem is reduced to a combinatorial optimization problem solved by using genetic algorithm. Orbits of toes and hip between stepping points are generated by means of parametric modeling. The stable walking patterns are obtained under the condition of the maximizing walking speed and the minimizing energy consumption. The Pareto front of the multi-objective optimization for the given robot model is visualized in advance by the MOGA, the optimum walking pattern is finally determined by using the satisficing trade-off method. The effectiveness of the proposed method is shown by simulation results.
Large-scale multiobjective design exploration using K computer is presented. This presentation includes aeroacoustic design optimization problem of a rocket launch site and aerodynamic design optimization problem of a plasma actuator applied to a wing model. Non-dominated solutions visualization tool named iSPM is also introduced.
"DESTINY" is an acronym of "Demonstration and Experiment of Space Technology for INterplanetary voYage", which is proposed by JAXA/ISAS as "ISAS Small Scientific Satellite" mission. In this mission, trajectory design is one an important technical element because of its many revolution low-thrust orbits with many mission objectives and constraints. Evolutionary computation is utilized to find candidates for the orbit.
The purpose of this study is to clarify the hydrodynamic vibration factors of automobile side-view mirror surface by using both methods experimental technique and CFD, which clarified the relationship between separated vortices which induce hydrodynamic forces and vibration of mirror surface. As a result, it was found that the frequency of the side-view mirror obtained by the vibration pickups is almost equivalent to that of vortices shedding from the mirror body itself. CFD made it clear that there exists the region with the largest force on the mirror surface where pressure fluctuations are induced by flow velocity fluctuations over the region.
An optimization approach is presented for design of a tuned mass damper called TD-TMD for three-directional seismic response reduction of structures. The mass damper consists of a viscous damper and a mass connected by flexible springs. By utilizing the flexibility of springs, the movement of the mass in three-directions and the elongation of viscous damper are amplified, and the vibration energy of the mass is effectively absorbed by the viscous damper. The TD-TMDs are attached to a latticed roof and its seismic responses are compared with those with conventional single-directional dampers. The objective function of the parameter optimization problem is the mean norm of the response displacements at the nodes of the roof. The bounds of parameters are determined by solving a auxiliary nonlinear programming problem to maximize the minimum deformation of the damper against static loads of various directions. The parameters are discretized into integer values, and approximate optimal solutions are found using a local search combined with pure random search that generates efficient intial solutions.
The present paper describes an application of non-parametric shape optimization to brake squeal phenomena. A main problem is defined as complex eigen value problem and a real part of the complex eigenvalue causing the brake squeal is chosen as an objective cost function. The Fre'chet derivative of the objective cost function with respect to the domain variation, which we call the shape derivative of the objective cost function, is evaluated using the solution of the main problem and the adjoint problem. To cope with numerical error of Fre'chet derivative caused by using component mode synthesis (CMS), we investigated the effect of the adoptive mode number on the complex eigen value. A scheme to solve the shape optimization problem is presented using an iterative algorithm based on the H^1 gradient method for reshaping. A numerical result of a detail brake model illustrates that the real part of the target complex eigenvalue monotonously decreases.
This paper presents a topology optimization method for minimizing the scattering of acoustic waves in an acoustic cloaking problem, based on a level set boundary expression and the Finite Element Method. We examine the dependence of two obtained acoustic cloaking designs on the frequency of the impinging sound waves. First, the topology optimization method using a level set model incorporating a fictitious interface energy for regularizing the topology optimization is discussed. Next, the optimal design problem for the acoustic cloaking is formulated. The constructed optimization algorithm uses the Finite Element Method for solving the wave propagation problem and the adjoint variable method for updating the level set function. The proposed method is applied to an enclosing cloaking problem and two numerical examples are provided, using 3000 Hz and 2500 Hz incident waves.
This paper presents industrial application of structural optimizations for automatic transmission of vehicles to achieve light weight and low radiated noise design. Optimization problems are formulated using types of size and shape optimization method and its purpose is to obtain optimal thickness distribution of the automatic transmission casing that minimizes mass and/or sound pressure radiated from automatic transmission. The optimization results show that the optimization method successfully found the optimization thickness distributions of automatic transmission casing that reduce mass and sound pressure.
Origin-Destination (OD) table is necessary to run microscopic traffic simulation but it is not observed directly. Therefore, we must estimate OD flow from observable information. It is a kind of the inverse problem. In this paper we present the iterative method to estimate OD flow by using link flow data derived from traffic censuses and some features of the method is checked by the simulation in simple road maps. Additionally, we modeled simplified traffic phenomena as a linear model for further studies. This model is based on superposition principle of traffic volume created from each OD flow and it may decrease calculation cost when counted to the iterative method.
In this paper, integrated optimization of SSTOs with two different engines systems was performed. As a result, although engine performances were different, similar airframe configurations and flight paths were obtained. In addition, different flight path with the similar airframe configuration was obtained as the temporary optimum solution in the optimization process. This transition of the optimum solution indicates the influence of SSTO's engine performance on its airframe configuration and flight path through the integrated optimization. Simultaneously, it was shown that improvement of the specific impulse in a low Mach number may improve the feasibility of a SSTO.
DESTINY is injected to long elliptical orbit by Epsilon rocket launcher. If the apogee altitude of the injected orbit is high enough, it is achieved to ease the requirements for design and operation of the spacecraft. This paper investigates the ability of trajectory injection by means of 4-stage Epsilon rocket using the method of multi objective optimization under several flight constraints.
In this research, a new method of optimizing component layout and fastening methods inside a product is developed for the facilitation of reuse and recycle. Due to rise of environmental awareness and enactment of legislation in recent years, products that reach their end-of-life need to be collected, disassembled and reused / recycled. However, since only high value components are reused / recycled and the rest is discarded from a cost effective standpoint, they need to be easily removed. In addition, since some components require regular maintenance, they also need to be easily removed / accessed. Therefore, the proposed method explores component layout and fastening methods in which their components can be removed with minimum disassembly time and effort by executing layout optimization and fastening method optimization cooperatively. In the case study, the proposed method is applied to design of internal devices of a laptop computer and the result is compared with the existing laptop computer.
In rubber extrusion processes, it is difficult to estimate the finish shape of products before actual molding dies are tested. Therefore, the development of molding dies needs a lot of die design, fabrication, and test processes of trial and error. In this study, an effective method for reducing the process is proposed. First, some design parameters for an important part of die shape are defined. Then, nine kinds of dies, which have varied values of the parameters, are designed with the experimental design method, and their prototypes are tested. Finally, the optimal die is derived from the test results by using the recsat evaluation method. The present approach was applied to an actual development process and discussed.
A mixed-integer linear programming method utilizing the hierarchical relationship between design and operation variables proposed to solve the optimal design problem of energy supply systems efficiently is extended to search K-best solutions: At the upper level, the optimal values of design variables are searched by the branch and bound method with operation variables relaxed to continuous ones; At the lower level, the values of operation variables are optimized independently at the respective periods set for variations in energy demands by the branch and bound method with the values of design variables given tentatively during the search at the upper level; The obtained solution is used to renew K-best incumbent solutions, and the upper bound for the value of the objective function for K-best solutions is replaced with the largest value of the objective function among K-best incumbent solutions. This method is implemented into a commercial solver. A practical case study on the optimal design of a cogeneration system is conducted, and the validity and effectiveness of the proposed method are clarified.
A method of evaluating the performance robustness of energy supply systems under uncertain energy demands is proposed based on the maximum regret criterion. The regret in the performance of a system with its design specified is evaluated with the performance of the system with its design optimized as a reference. The regret is maximized with respect to uncertain energy demands, and this optimization problem is formulated as a multilevel mixed-integer linear programming one. At the first step, uncertain energy demands are discretized, the energy demands which maximize the regret are selected among their discretized values, which means that the maximum regret is underestimated, or the robustness is overestimated. In addition, a mixed-integer linear programming method utilizing the hierarchical relationship between design and operation variables proposed previously is adopted to solve the optimization problem efficiently. A case study on a gas turbine cogeneration system for district energy supply is conducted, and the maximum regret in the annual total cost of a system is evaluated. Through the study, the validity and effectiveness of the evaluation method and features of the economic robustness of the system are clarified.