Ichirou Ishimaru

A proposal of automatic generation of fuzzy neural network and its application to precise adjustment system

Recently the need for automation of higher proficiency adjustment is increasing. However, it has taken an enormous amount of time to make the adjustment algorithm and timely development of the automation equipment has not happened. The purpose of the paper is to reduce the term for the development of the adjustment algorithm. The paper presents a construction method for fuzzy neural networks. We propose a method to select the input parameters that can recognize vague waveforms. To operate the learning method easily, we propose an automatic learning rate set-up method and also propose a method to avoid the local minimum automatically. We also propose a virtual experiment system using fuzzy neural networks to evaluate the ability of the learned algorithm. We have developed a system which integrates the proposed construction method of fuzzy neural networks and the virtual experiment system.

A new reformation scheme of steel bars and a learning method for various plastic characteristic using fuzzy clustering

In order to secure the precise straightness, the steel bars is reformed by bender. Up until now, this bender has been operated by experts. However, due to the shortage of experts in this field, an automatic reformation process is required. This paper presents a reformation scheme to cope with variations in physical characteristics of the steel bars, and an estimation method of the stress required to reform the steel bars using a fuzzy clustering method. The precision of the reform experiment is shown to meet the requirements.

6-DOF Position and Attitude Measurement Technology by Diffracted Light Probe

In this paper, we proposed a six-degree-of-freedom (6-DOF) measurement technology using the first order diffracted light as optical probe which is called “diffracted light probe”. The position of the diffracted light probe which is detected by the CCD camera changes according to the position and attitude of the grating. So, by calculating the optical axis of diffracted light probe, we can know the position and attitude of the grating simultaneously. In this method, the intensity distribution of the diffracted light affects the measuring accuracy. So, we established the design procedure based on the grating equation and by using the annular filter. Based on this procedure, we formed the grating (width of slit: 2μm, span of slits: 10μm, number of slits: 20). In addition, we formed the annular filter whose gap is 200μm by RIE-ICP.By introducing these products into the optical system, we could obtain the diffracted light probe for the high precision measurement and verified the principle of this proposed method with the good results (position accuracy: 8μm, attitude accuracy: 0.2deg.).

Measurement Technology for Relative Position and Attitude of Micro-manipulator

近年, MEMS技術は様々な高機能微小構造体を提供することが可能となっており, これらの高機能微小構造物をウェハから取り出して機能部品として活用することができれば, さらに応用範囲を拡張することが期待できる. そこで, MEMSを機能部品として扱うためのMEMSの位置姿勢同時計測技術の確立が重要であると考えている. そこで, 我々はMEMSの相対位置姿勢計測技術として近接共焦点位置姿勢同時計測手法を提案した. 本報では, 提案手法の原理確認及び精度検証を行い, その実用性について検証を行った.

A proposal of probability type fuzzy inference and its application to precision plastic deformation algorithm

Proficiency in creating a knowledge base is required for high accuracy fuzzy control. To achieve this we propose a new fuzzy inference method utilizing probability density functions of scattering of measurement values defined as membership functions. In addition, to consider qualitative knowledge of the data distribution, we propose a method to determine the membership functions using the distributions of grade accumulation. We apply the proposed method to automatic high accuracy adjustment devices for magnetic heads.

A development of an automatic learning method for various plastic characteristic using fuzzy clustering and its application

We have proposed a new plastic deformation control method (Ishimaru et al. (1999)) based on work hardening. This method calculates the deformation load in accordance with average plastic characteristic such as n-value and work hardening, but drawback is that the adaptability for each work piece is not satisfactory. These mechanical characteristics of steel bar are a little different in each workpiece. In this case, the adaptability of the control algorithm for the change of the applied products characteristics is a very important factor. In this paper, a new online learning algorithm utilizing fuzzy clustering is proposed. This new learning algorithm can realize adaptability for the difference of plastic characteristics of each workpiece by deriving and analyzing the abundant data obtained automatically in a mass production line.