著者らは多次元データ可視化手法の一環として，多次元データから生成可能な多数の散布図のうち閲覧価値がありそうな散布図群を自動選出する手法を提案している．この手法が採用している散布図選出基準の中には，多次元データ中のカテゴリ型変数で散布図中の点群を色分けした結果を対象とする基準がある．本報告ではこの手法を拡張して，多次元データ中の特定の変数の大小で散布図中の点群を色分けして可視化した結果を示す．この拡張手法は，多次元データ中の特定の変数を色で表現し，それ以外の任意の 2 つの変数を 2軸に割り当てた散布図を生成することで，実質的に 3 つの変数を表現する多数の散布図の中から閲覧価値がありそうな散布図群を自動選出していることに相当する．

抄録全体を表示

There exists the immune algorithm as a computational method in solving optimization problems. This algorithm imitating the immune mechanism in the mammal has a notable characteristic that is able to search a solution set with diversity in optimization problems. However, application of the immune algorithm to structural optimization problems is a few in the present stage. In this paper we would like to improve the conventional immune algorithm with the gene of binary number type to immune algorithm with the gene of a binary-coded type and a real-coded type in applying to structural optimization. The applicability and effectiveness of our immune algorithm is shown through numerical models on the minimum weight design and the topology specification of elastic truss structure. The obtained numerical results are compared with the results by the standard genetic algorithms and the real-coded genetic algorithms. Moreover, in our immune algorithms an efficient procedure on selection of four computation thresholds in order to generate a solution set with diversity is also proposed.

抄録全体を表示

GAは離散的最適化問題の解法として有用である.著者らも交配個体選択GAと名付けた独自のGAを提案し,その解法特性を検討してきた.しかし混合型最適化問題に対するGAによる解法に関する研究はあまりない.本研究では,離散変数は従来の2進数線列で構成される離散型遺伝子によって表現し,遺伝子によって連続変数をそのまま表現することにより,種々の操作を必要としないままに混合型最適化問題を解く方法を提示し,その解法特性を基礎的に検討する.

抄録全体を表示

In this paper, we discuss the optimization of the high-pressure-water washing system used in the recycling process for polystyrene-paper(PSP) trays used for wrapping fishes, which is expected to reduce the cost of recycling and environmental load. The washing time and energy required to clean a test tray were measured under various water pressures and flow rates. Two criteria for optimization in terms of energy saving and cost saving were developed on the basis of the measurement results, the respective optimum washing conditions were then calculated. Moreover, we proposed a method of searching for the optimum condition that gives the lowest cost under a given allowable energy consumption.

抄録全体を表示

While double-precision real numbers have been used for CAE simulations based on the finite element method, low-precision real numbers are used for deep learning for achieving much faster learning and inference. Due to overwhelming demand for deep learning, low-precision arithmetic is beginning to be introduced into general-purpose processors as accelerators. By using low-precision real arithmetic for finite element analysis, we can effectively use the arithmetic unit for deep learning and get advantages in terms of calculation speed and power consumption. In this study, in order to investigate the applicability of low-precision computation to finite element analysis, low-precision computation is firstly applied to the element integration process, and its effects are quantitatively evaluated.

抄録全体を表示

For CAE simulations based on the finite element method, double-precision real numbers, which requires more energy than others, have been almost exclusively used. On the other hand, due to overwhelming demand for deep learning, low-precision arithmetic is now being introduced into general-purpose processors as accelerators. With SDGs (Sustainable Development Goals) in mind, it is apparent that we must save energy as possible as we can. By using low-precision real arithmetic for finite element analysis, we can effectively use the arithmetic unit equipped for deep learning and get advantages in terms of calculation speed and power consumption. In this study, a new energy-saving finite element analysis assisted by deep learning is proposed. In the proposed method, deep learning indicates where to use lower precision computation for saving energy. Here, for the first try, the effect of low precision numbers on analysis results is quantitatively investigated through sample analyses.

抄録全体を表示

一般的な離散ウェーブレット変換におけるシフト不変性の欠如は,これらの変換を信号処理や画像処理に利用する際に,しばしば深刻な問題となることが指摘されている.本論文では,このシフト不変性の欠如の問題を周波数領域において考察する際に,有用となる定理を証明する.この定理は任意の2乗可積分複素数値関数の複素数離散ウェーブレット変換における,完全シフト不変性を保証する条件を与える完全シフト不変定理を,より一般化したものである.次にこの定理を用いて,完全シフト不変性を実現するタイトウェーブレットフレームを提案する.

抄録全体を表示

進化計算の分野において，繰り返し囚人のジレンマ (IPD) ゲームは，エージェントの協調行動の進化の例として盛んに研究されている．本研究では，トーラス状の２次元格子空間上にプレイヤを配置する．各プレイヤは，バイナリ型かの戦略表現形式を持つものとする．それぞれの表現形式は戦略進化に関して異なる特徴を持つ．本デモンストレーションでは，異なる戦略表現を持つプレイヤの配置の違いが協調行動の進化に与える影響を示す．例えば，同一の戦略を持つプレイヤを一箇所に配置した場合と，戦略の表現型に関係なくランダムに配置した場合を比較する．協調行動の進化の過程が視覚的に分かるように開発したインタフェースを用いてデモンストレーションを行う．

抄録全体を表示

連続ウェーブレット変換(Continuous Wavelet Transform; CWT)は時間-周波数解析の道具として優れた性質をもち,さまざまな分野での応用が試みられているが,変換の核となるマザーウェーブレット(Mother Wavelet; MW)の選択が重要であるにもかかわらす,統一された選択方法がないのが現状である.本研究ではモデル信号を用いてMWの特性を比較検討し,それらの選択について検討した.その結果,通常の時間-周波数解析の場合,複素数型MWを選択する方がよいという知見が得られた.また,MWはその関数の対称性が重要な個性であり,信号の中の異なる性質を抽出するときなどに有効であった.

抄録全体を表示

In most CAE simulations based on the finite element method, double-precision real numbers have been used. On the other hand, due to overwhelming demand for deep learning, low-precision arithmetic is now being introduced into general-purpose processors. By using low-precision real arithmetic for finite element analyses, we can effectively use the arithmetic unit originally equipped for deep learning and obtain advantages in terms of calculation speed. In this study, we propose a fast finite element analysis assisted by deep learning, where low-precision arithmetic is aggressively used to accelerate calculation speed. In the proposed method, deep learning indicates where to use lower precision computation for calculation speed. Here, for the first try, the effect of low precision numbers on numerical quadrature of finite elements is quantitatively investigated through sample analyses.

抄録全体を表示

コンピュータ計算がわれわれの研究で大きな場所を占め, 乱数がしばしば使用されるようになってきた. 大抵の場合, コンピュータに組込まれている算術乱数でこと足りるのであるが, 大量の乱数を必要とするときや多次元の計算をするときなど, これが擬似乱数であるため, いろいろと心配がでてくる. 乱数計算をしていて思いもかけぬ結果がでてきて当惑した経験を持たれている方もおられるだろう. 良い乱数, 安心して使える乱数はないものだろうか. それには, われわれがランダムであると信じて疑わない物理現象を利用して乱数を発生するのが一番よい. こうして作った物理乱数がようやく実用に供せられるところにきたのである.

抄録全体を表示

With the development of the modern vehicle industry, research and development of intelligent transport system has been undertaken. As an active safety feature, there is a need to develop a system with an appropriate steering and braking operation for accident avoidance. In this study, it is assumed that an automatic vehicle, which runs on the highway, encounters an obstacle in distance of collision by braking without steering operation. Our study aims to find operations that the vehicle can perform to avoid an obstacle by a combination of steering and braking. This paper proposes a method to generate operating waveforms using an immune algorithm (IA) that can generate plural quasi-optimum solutions. CarSim (Mechanical Simulation Company), which possesses full vehicle dynamics with 27 degrees of freedom, is adopted as the vehicle model in our study. Inputs for this model include wheel steering angle and brake pressure. Operation waveforms are generated by linear interpolation through designated data points. Real number type IA, which is coded real number expression, is adopted in order to find data points. The efficacy of the proposed method is verified through an emergency avoidance simulation using CarSim. The simulation result demonstrates operation patterns to avoid an obstacle; in addition, real number type IA is shown to perform better than conventional binary type in a search performance of the solution.

抄録全体を表示

We have been studying complex phenomena on flows of vesicles suspending to blood in micro circulations. In this paper, the distribution of vesicles at an orthogonal bifurcation is investigated in terms of numerical simulations by using real-coded lattice gas model, and we study its dependency on both the concentration of vesicles in the mother vessel and the diameter-ratio between two daughter vessels. Focusing on the ratio between the number of vesicles at each bifurcated vessel, it obeys the function of the third power of diameter-ratio of two daughters except the small diameter-ratio case. This might indicate that the flow ratio between two daughters would be the main physics on the dividing concentration of vesicles at the bifurcation.

抄録全体を表示

進化戦略 (Evolution Strategy: ES) は, 大域的最適化手法の一つである.ESのアルゴリズムは, 多点探索や交叉が導入されている点ではGAに類似している.しかし, ESの主な特徴は, 決定変数に正規乱数の摂動を加える形で突然変異を行い, これを主な探索手段としていること, 摂動の大きさが自己適応によって変化することにある.本研究では, ESをタンクモデル定数の同定に適用し, 真値を設定した数値実験によってESの探索能力を検証した.また, 永源寺, 大迫, 青蓮寺ダム流域の観測流量資料によるタンクモデル定数の同定を試みた.これらの検討により, ESおよび並列型ESの探索能力は, バイナリGAおよびGAよりはるかに優れており, 得られたモデル定数の一貫性に関しては, 並列型ESは, 現在のところ最も強力な最適化手法と目されているSCE-UA法と同等ないしより強力であることが示された.

抄録全体を表示

This paper presents a redesign of the conventional vaned diffuser into the low solidity diffuser for the maximum static pressure recovery for diffuser pump. A Bezier curve representation for profile description is coupled with a blade-to-blade flow calculation and a real-coded Genetic Algorithm. A low solidity vaned diffuser of 0.89 in solidity is obtained through the present optimization design method. Numerical analysis and experimental test are applied to evaluate the hydrodynamic performance of diffuser pump with the designed low solidity vaned diffuser and original vaned diffuser. The obtained results demonstrate that the diffuser pump with the optimized low solidity diffuser has compact size compared with the original one while ensuring the performance requirements.

抄録全体を表示