3D Precise Inspection of Electronic Devices by

3 downloads 0 Views 871KB Size Report
from the observable direction “A”. Therefore, the system which performs 3D measurement and measures the dis- tance is required. Figure 2. Terminal lead with ...
13-5

MVA2007 IAPR Conference on Machine Vision Applications, May 16-18, 2007, Tokyo, JAPAN

3D Precise Inspection of Electronic Devices by Single Stereo Vision Takashi Watanabe School of Information Science and Technology, Chukyo University 101 Tokodachi, Kaizu-cho, Toyota, Aichi 470-0393 Japan E-mail: [email protected]

Akira Kusano Institute for Advanced Studies in Artificial Intelligence, Chukyo University 101 Tokodachi, Kaizu-cho, Toyota, Aichi 470-0393 Japan

Takayuki Fujiwara SIST, Chukyo University 101 Tokodachi, Kaizu-cho, Toyota, Aichi 470-0393 Japan E-mail: [email protected]

Hiroyasu Koshimizu SIST, Chukyo University 101 Tokodachi, Kaizu-cho, Toyota, Aichi 470-0393 Japan E-mail: [email protected] method are normally used for these 3D measurements.[7] However, the laser apparatus for lighting is needed and it will become an expensive and complicated system. To do this in our research, we propose an inspection environment for realizing not expensive and highly precise 3D measurement scheme and its algorithm.

Abstract It is very important to guarantee the quality of the industrial products by means of visual inspection. In order to reduce the soldering defect with terminal deformation and terminal burr in the manufacturing process, this paper proposes a 3D visual inspection system based on a stereo vision with single camera. It is technically noted that the base line of this single camera stereo was precisely calibrated by the image processing procedure. Also to extract the measuring point coordinates for computing disparity; the error is reduced with original algorithm. Comparing its performance with that of human inspection using industrial microscope, the proposed 3D inspection could be an alternative in precision and in processing cost. Since the practical specification in 3D precision is less than 0.02 mm and the experimental performance was around the same, it was demonstrated by the proposed system that the soldering defect with terminal deformation and terminal burr in inspection, especially in 3D inspection, was decreased. In order to realize the inline inspection, this paper will suggest how the human inspection of the products could be modeled and be implemented by the computer system especially in manufacturing process.

1.

2.

The necessity for 3D measurement

Figure 1 show the soldering state of a terminal lead, which consists of copper alloy. Since the size of the device and the number of terminals are not always fixed, inspection system must cope with the variations both in size and in number, and must cover 3D measurement supplemental to 2D measurement. 㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻

㪧㪚㪙

㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻

㪧㪚㪙

Figure 1. Soldering state of a terminal lead Since the 2D measurement of the terminal lead with the deformation or terminal cutting burr that is difficult to extract only from the appearance information, it is necessary to measure the distance between the camera and a terminal’s measurement point by stereo system. In order to detect simultaneously the deformation, bend, garbage adhesion and terminal burr as shown in Figure 2, 3D measurement is essential. The appearance of the defect of the terminal lead cannot be recognized from the direction shown as “䎥” in Figure 2. It is necessary to measure the 3D shape using the image information seen from the observable direction “䎤”. Therefore, the system which performs 3D measurement and measures the distance is required.

Introduction

It is indispensable to prepare for the spreading variations in product design and for the shortage in product life cycle, under the changing circumstances of portable phone, PC, and several electronic devices. Accordingly the inspection of the electronic devices must be shorten in development period, must be enforced in precision and also be refined in size.[1],[2],[3] Therefore image processing is becoming one of the  important  key technologies in this field. This paper proposes a flexible image processing system for the inspections of electronic device's terminal lead such as small size connectors implemented on general purpose PC. [4],[5],[6] In the real production line, inspection of electronic devices is categorized into two basic categories: visual inspection and size measurement. The former is a task for replacing human visual inspection and the latter is one for replacing human works using special measurement equipment. In order to introduce 3D measurement for the 3D form of the electronic devices of which form is complicated, and to introduce high precision measurement, an optical probe system or the optical cutting

Figure 2. Terminal lead withsome problems 449

3.

4.2.

Proposed method

Accuracy is extremely important in the extraction of a measuring point in this system. An accurate measurement of the reference point of the device must be realized from a pair of stereo images in order to measure the real length of the movement of the feeder as the base line length.

In order to realize the function of stereo views with single camera, a mechanism shown in Figure 3 was fabricated, where a camera is fixed and the work is set movable along with the feeder, and the left and right stereo images were captured at the respective moment before and after the mechanical feeder motion. The base line was precisely calibrated by using the corresponding feature points extracted from these two images. 㪪㫐㫄㪹㫆㫃㩷㫄㪸㫉㫂 㪙㫃㫆㪺㫂 㪮㫆㫉㫂

Relative stabilization of the feeder motion For providing the accurate length of the base line, it is necessary to compensate the position variations of the mechanical control of the feeder due to the inertia of the feeder. Then we introduced the following countermeasure based on an image processing method applicable to a set of stereo images: 1) First let the symbol mark of the block stage be extracted. 2) Then let the coordinate system of the respective device be generated at the mark as an origin. 3) Step 1) and 2) are repeated at every motion of the feeder. 4) Let the difference of a pair of origins be calculated as the length of the base line.

㪤㫆㫍㪼㩷㪻㫀㫊㫋㪸㫅㪺㪼㪔 㪙㪸㫊㪼㫃㫀㫅㪼㩷㫃㪼㫅㪾㫋㪿㩷㩹㪛㩹

㪝㪼㪸㫋㫌㫉㪼㩷㫇㫆㫀㫅㫋 㩿㪹㪼㪽㫆㫉㪼㩷㫄㫆㫍㪼㫄㪼㫅㫋㪀



㪝㪼㪸㫋㫌㫉㪼㩷㫇㫆㫀㫅㫋 㩿㪸㪽㫋㪼㫉㩷㫄㫆㫍㪼㫄㪼㫅㫋㪀 㪚㪸㫄㪼㫉㪸



Measurement points extraction Feature points are extracted from a pair of stereo images for providing the stereo correspondence by the following procedure: 1) A small local region is prepared for the intensive image processing based on the origin preliminary extracted. 2) Edge image is extracted from the small local region by using Laplacian-Gaussian processing. 3) In order to extract the local coordinate system on the device, a pair of perpendicular straight lines shown in Figure 5 is extracted from the edge image by RMS line fitting procedure. 4) Let the crossing point of these fitted lines be the origin of the local coordinate system of this device. 5) The circle of arbitrary radius and the intersection of RMS line centering on the origin are considered as the position of the feature point extraction (Figure 5). 6) All edge points are represented again based on this local coordinate system.        

Figure 3. Single camera stereo model Primal feature points can be extracted from the upper surface of the terminal lead, and these feature points can be utilized to detect the defects by measuring the respective distance between the point (hԘ, hԙ, …, hԛ in Figure 4) and the camera along the edge of the terminal lead. 㪚㪸㫄㪼㫉㪸

㪟㽲

㪟㽳

㪟㽴

㪟㽵

㪤㪼㪸㫊㫌㫉㪼㫄㪼㫅㫋㩷㫇㫆㫀㫅㫋

㪢 㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻 䌨㽳

The method of measuring point extraction

䌨㽵 䌨㽴

䌨㽲 㪚㫃㪼㪸㫉㪸㫅㪺㪼

㪙㪸㫊㪼㩷㪹㫃㫆㪺㫂

㪩㪸㪻㫀㫌㫊㩷㩹㫉㩹

Figure 4. Inspection method model

4.

㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻

㪱㫆㫆㫄

The unstable element and measurement method of inspection

㪜㪻㪾㪼㩷㫃㫀㫅㪼

㪦㫉㫀㪾㫀㫅㩷㫇㫆㫀㫅㫋㩷㩹㫈㩹

㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻

㪤㪼㪸㫊㫌㫉㫀㫅㪾 㫇㫆㫀㫅㫋㩷㩹㫇㩹

4.1. Why base line must be concurrently adjusted

㪝㫀㫋㫋㫀㫅㪾㩷㫊㫋㫉㪸㫀㪾㪿㫋 㫃㫀㫅㪼㩷㩿㫐㩷㪻㫀㫉㪼㪺㫋㫀㫆㫅㪀

At every time when the device feeder stops around in front of the camera, the feeded distance must be exactly measured, because the mechanical feeder does not perform so sufficient in precision that the base line length can be estimated. In addition, the device on the feeder is likely to fluctuate both horizontally and vertically. In order to cope with these situations, we installed a hardware equipment not to suffer from the mechanical vibration, and we developed a software countermeasure to estimate the exact base line length by the image processing technique. The details of the algorithm will be presented in the next chapter.

㪝㫀㫋㫋㫀㫅㪾㩷㫊㫋㫉㪸㫀㪾㪿㫋 㫃㫀㫅㪼㩷㩿㫏㩷㪻㫀㫉㪼㪺㫋㫀㫆㫅㪀

㪠㫅㫋㪼㫉㫊㪼㪺㫋㫀㫆㫅 㩷㩿㪦㫉㫀㪾㫀㫅㩷㫇㫆㫀㫅㫋㩷㩹㫈㩹㪀

Figure 5. The measuring point extraction

4.3.

Stereo correspondence algorithm

Affine transform for rotation Figure 6 shows an algorithm for compensating the rotation fluctuation of the device, and Figure 7 and Figure 8 shows the detailed definition of the origin O of the co-ordinate system. 450

When the origin on the measurement point p(x,y) moves to p'(x',y'), the correspondence between these p and p' is formalized by eq.'s (1), (2) and (3) where the angle ș1 and ș2 are the rotation and R is the radius of the rotation. 'c R cosT 2  R cosT1 (1) 'k R sin T 2  R sin T 1 (2) § x' · p c¨¨ ¸¸ © y' ¹

§ x · § 'c · ¨¨ ¸¸  ¨¨ ¸¸ © y ¹ © 'k ¹

It is because this system is manufactured. H h

KD d

(4)

K  H (5) 㪙㪸㫊㪼㫃㫀㫅㪼㩷㫃㪼㫅㪾㫋㪿 㪛㩿㫄㫄㪀 㫑

㪪㫐㫄㪹㫆㫃㩷㫄㪸㫉㫂 㪙㪸㫊㪼㪹㫃㫆㪺㫂 㫊㫌㫉㪽㪸㪺㪼

(3) 㪢



㪧㪈 㪧㪉



㪥㫌㫄㪹㪼㫉㪄㫆㪽㪄㫋㫀㫄㪼㫊㩷㫉㪼㫊㪼㫋㩷㫆㪽㩷㫇㫉㫆㪺㪼㫊㫊㫀㫅㪾㪅㩿n 㪔㪇㪀

㩿㫏㪃㫐㪀㪔㩿㪇㪃㪇㪀

㪫㪿㪼㩷㫊㫋㫉㪸㫀㪾㪿㫋㩷㫃㫀㫅㪼㩷㫀㫅㪾㫉㪼㪻㫀㪼㫅㫋㩷㫆㪽㩷㪸㩷㫎㫆㫉㫂㩷㫀㫊㩷㪼㫏㫋㫉㪸㪺㫋㪼㪻㪅

㫇㪉

㫐 㫇㪈

㪛㫀㫊㫇㪸㫉㫀㫋㫐 㪻㩿㫄㫄㪀

n=n+1

㪝㫆㪺㪸㫃 㫃㪼㫅㪾㫋㪿 㫏

㪠㫄㪸㪾㪼

㪫㪿㪼㩷㪸㫅㪾㫃㪼㩿șn 㪀㩷㪹㪼㫋㫎㪼㪼㫅㩷㫋㪿㪼㩷㪼㫏㫋㫉㪸㪺㫋㪼㪻㩷㫊㫋㫉㪸㫀㪾㪿㫋㩷㫃㫀㫅㪼 㪸㫅㪻㩷㫇㪸㫉㪸㫃㫃㪼㫃㩷㫃㫀㫅㪼㫊㩷㫀㫊㩷㪺㪸㫃㪺㫌㫃㪸㫋㪼㪻㪅㩷㩷㩿n 㪔㩷㪈㪃㩷㪉㪀

Figure 9. Triangular principle for stereo system Inspection of terminal lead As shown in Figure 10, one terminal lead has four measuring points and the respective height of those measuring points is measured as height hԘ - hԛ. And as compared with the threshold th, the success and rejected judgments are carried out for the computed height of those four points given by eq. (6).

No n=1 Yes 㪘㩷㫎㫆㫉㫂㩷㫀㫊㩷㫄㫆㫍㪼㪻㪅

‫ޓ‬hԙ | hԚ | hԛ : hԘ : deformation ‫ޓ‬hԘ | hԛ  hԙ | hԚ : bend ‫ޓ‬hԘ  hԙ | hԚ | hԛ : ter min al _ burr

㪫㪿㪼㩷㫍㫀㫉㫋㫌㪸㫃㩷㫊㫋㫉㪸㫀㪾㪿㫋㩷㫃㫀㫅㪼㩷㫎㪿㫀㪺㪿㩷㪺㫆㫅㫅㪼㪺㫋㪼㪻㩷㫋㪿㪼 㫄㪼㪸㫊㫌㫉㫀㫅㪾㩷㫇㫆㫀㫅㫋㩷㩿p 㪀㩷㪽㫉㫆㫄㩷㫋㪿㪼㩷㫉㫆㫋㪸㫋㫀㫆㫅㩷㪺㪼㫅㫋㪼㫉㩷㩿O 㪀

‫ޓ‬th  hԘ | hԙ | hԚ | hԛ : micro _ garbage (6)

㪛㫀㫊㫋㪸㫅㪺㪼㩷㪹㪼㫋㫎㪼㪼㫅㩷㩿O 㪀㩷㪸㫅㪻㩷㩿p 㪀㩷㫀㫊㩷㫊㪼㫋㩷㫋㫆㩷㩿R 㪀㪅 㪫㪿㪼㩷㫍㫀㫉㫋㫌㪸㫃㩷㪺㫀㫉㪺㫃㪼㩷㫎㪿㫀㪺㪿㩷㫄㪸㪻㪼㩷㫋㪿㪼㩷㫉㪸㪻㫀㫌㫊㩷㩿R 㪀㩷㫎㪿㫀㪺㪿 㫄㪸㪻㪼㩷㩿O 㪀㩷㫋㪿㪼㩷㪺㪼㫅㫋㫉㪸㫃㩷㫇㫆㫀㫅㫋㩷㫀㫊㩷㪾㪼㫅㪼㫉㪸㫋㪼㪻㪅 㪘㫃㫆㫅㪾㩷㫎㫀㫋㪿㩷㪸㩷㫍㫀㫉㫋㫌㪸㫃㩷㪺㫀㫉㪺㫃㪼㪃㩷㪸㩷㫄㪼㪸㫊㫌㫉㫀㫅㪾㩷㫇㫆㫀㫅㫋㩷㩿p 㪀㩷㫀㫊 㫄㫆㫍㪼㪻㩷㪹㫐㩷㪸㫅㪾㫃㪼㩷ș 1 㫋㫆㩷ș 2 㪅

㪇㪅㪍㪌

㪫㪿㪼㩷㪺㫆㫆㫉㪻㫀㫅㪸㫋㪼㫊㩷㫆㪽㩷㫋㪿㪼㩷㫄㪼㪸㫊㫌㫉㫀㫅㪾㩷㫇㫆㫀㫅㫋㩷㪸㪽㫋㪼㫉 㫄㫆㫍㪼㫄㪼㫅㫋㩷㩿p 㪀㩷㪸㫉㪼㩷㪼㫏㫋㫉㪸㪺㫋㪼㪻㪅㩷㩷䈀㩿p 㪀㸢㩿p 㩾㪀䈁

㪇㪅㪋 㪇㪅㪊 㪇㪅㪉 㪇㪅㪈

Figure 6. The flow of rectifies posture change of a work 㪤㪼㪸㫊㫌㫉㫀㫅㪾㩷㫇㫆㫀㫅㫋㩷㩹㫇㩹

㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻









㪩㫆㫋㪸㫋㫀㫆㫅㩷㪺㪼㫅㫋㪼㫉 㫇㫆㫀㫅㫋㩷㩹㪦㩹

㪿㽲

㫇㩿㫏㪃㫐㪀

㪿㽳

㪿㽵 㪿㽴

Figure 10. Terminal lead measurement points 㪩

5. 㪦

Figure 7. Rotation center point extraction 㫇㩿㫏㪃㫐㪀

㼺㪺

㼺㫂

㫇㩾㩿㫏㩾㪃㫐㩾㪀

5.1.

Performance evaluation of inspection system Common equipments

The inspection system was implemented on a PC with the following specifications: 1) PC processor: celeron 2.4GHz 2) OS: Windows2000 3) Software: HALCON (MV-Tec, LinX)Visual C++ (Microsoft) 4) Memory: from 100 to 150MB

㱔㪈 㱔㪉 㫏

Figure 8. The measuring point after rotation compensation

5.2.

Stereo measurement method Stereo measurement shown in Figure 9 is implemented based on the triangular principle given by eq.’s (4) and (5). The heights H and h of a feature point can be calculated from the disparity d and baseline length D. However we know the distance K between the base block and the camera.

Processing time

The processing time was 0.65sec in average for one device (40 terminals).

5.3.

451

Experimental performance evaluation

Reliability of work posture change In order to check the change of the measurement value hԘ, the amount of posture rotations of one sample is changed from 0.31degrees to 1.58degrees (Figure 8: ș2). When the tolerance level of the change was set up with the picture resolution of 0.03mm, it became clear that the amount of change became to be 1.30degrees (Figure 14).

Inspection result of terminal lead The electronic devices of Figure11 with 40terminals were able to be measured, and the detailed shape change of a terminal lead has been recognized. Figure 11 shows the graph of the terminal lead measurement result of four samples. 㪥㫆㪅㪈 㪥㫆㪅㪉

㪥㫆㪅㪊㪐 㪥㫆㪅㪋㪇

㪫㪼㫉㫄㫀㫅㪸㫃㩷㫃㪼㪸㪻

㪂㪇㪅㪇㪊㫄㫄 㪙㪸㫊㫀㪺㩷㪻㪸㫋㪸 㩿㱔䋲㪔㪇㪅㪊㪈㫦㪀

㱔䋲

㪇㪅㪈

㪿㽲

㪥㫆㪅㪈 㪈

㪇㪅㪇㪉㪌

㪇㪅㪇㪉㪊

㪇㪅㪇㪉㪊

㪇㪅㪇㪉㪈

㪇㪅㪇㪉㪈

㪇㪅㪇㪈㪐 㪿㽲

㪇㪅㪇㪈㪐

㪿㽳

㪿㽴

㪪㪈

㪿㽵

㪄㪇㪅㪇㪊㫄㫄

㪇㪅㪇㪈㪌

6.

㪿㽲㩷㩿㫄㫄㪀

㪇㪅㪇㪋 㪇㪅㪇㪊 㪇㪅㪇㪉 㪇㪅㪇㪈 㪇 㪐

㪈㪈 㪈㪊 㪈㪌 㪈㪎 㪈㪐 㪉㪈 㪉㪊 㪉㪌 㪉㪎 㪉㪐 㪊㪈 㪊㪊 㪊㪌 㪊㪎 㪊㪐

㪫㪼㫉㫄㫀㫅㪸㫃㩷㪥㫆㪅

Figure 12. Deviation in measurements of the proposed system Reliability of work movement distance In order to check the change of the measurement value hԘ, the movement distance of one sample is changed from 30mm to 25mm. When the tolerance level of the change was set up with the picture resolution of 0.03mm, it became clear that the amount of change became to be 25.5mm (-4.5mm)(Figure 13). Even when the movement distance of the work is changed more, the measurement accuracy was maintained. It means that the image capturing and the inspection of it are successfully achieved during the work movement. 㪙㪸㫊㫀㪺㩷㪻㪸㫋㪸 㩿㪛㩷㪔㩷㪊㪇㫄㫄㪀

㪊㪇㫄㫄

㪉㪐㫄㫄

㪉㪏㪅㪌㫄㫄

㪉㪏㫄㫄

㪉㪎㪅㪌㫄㫄

㪉㪎㫄㫄

㪉㪍㪅㪌㫄㫄

㪉㪍㫄㫄

㪉㪌㪅㪌㫄㫄

㪉㪌㫄㫄

㪉㪈

㪉㪌 㪉㪎

㪄㪇㪅㪇㪊㫄㫄

㪿㽲㩷㩿㫄㫄㪀

㪇㪅㪇㪍 㪇㪅㪇㪋 㪇㪅㪇㪉 㪇

㪛㪔㪉㪌㫄㫄

㪄㪇㪅㪇㪋 㪈









㪈㪈

㪈㪊

㪈㪌

㪈㪎

㪈㪐

㪉㪊

㪉㪐



㪈㪈 㪈㪊 㪈㪌 㪈㪎 㪈㪐 㪉㪈 㪉㪊 㪉㪌 㪉㪎 㪉㪐 㪊㪈 㪊㪊 㪊㪌 㪊㪎 㪊㪐

㪊㪈

㪊㪊

㪊㪌 㪊㪎

Conclusion and future subjects

[1] H.Tamura: Computer Image Processing, Ohmsha(2002) [2] M.Ejiri: Industrial applications of image processing (Examples), FUJI techno system (1994) [3] H.Koshimizu: Today’s automated visual inspection systems, Techno system (1990) [4] Linx Corp.: Practice image processing, Springer -Verlag Tokyo (2000) [5] T.Watanabe, H.Koshimizu: Image Processing Algorithm for Alignment Quality Control of Molded Electronic Devices Datum Surface of Printed Board, IEEJ Trans. EIS,Vol.124,No.3, pp.740-747 (Mar.2004) [6] T.Watanabe, A.Kusano, T.Fujiwara H.Koshimizu: A Proposal of 3D Measuring Method by Means of A Single Camera and Its Application to Precise Baseline Detection Algorithm for Electronic Device Inspection, Proc.IWAIT2006, pp.42-47 (Jan. 2006) (Naha, Japan) [7] T.Yoshizawa: Optical 3D measurement technologies, Shingijyutsu communications (1993)

㪇㪅㪇㪏

㪄㪇㪅㪇㪉



References

㪙㪸㫊㪼㫃㫀㫅㪼㩷㫃㪼㫅㪾㫋㪿㩷㩿㪛㪀

㪂㪇㪅㪇㪊㫄㫄



We evaluated the measurement accuracy and the adaptability to the posture change of a work by the experimental equipment. The processing time was 0.65sec for one device. This processing time was estimated based on 50times experiments. And from the result of an experiment, when the baseline length changes with a maximum of 4.5 mm or 1.3 degrees of the device fluctuation in motion, it was clarified experimentally that the error in the positioning procedure was reduced to 0.02mm or less. Through these technical discussions, we could demonstrate that a high precision measurement system for small electronic devices can be developed even if the devices are fed without specialized mechanical stabilization both against the horizontal positioning and against the vertical regulation in posture. In order to improve the performance and to put the system more practical, the followings should be solved: 1) A long term field test at the real production line should be executed to evaluate and enforce the performance. 2) It is also expected to make the proposed system to be applicable to the wider scope of the similar but different electronic devices.

㪇㪅㪇㪌





㪫㪼㫉㫄㫀㫅㪸㫃㩷㪥㫆㪅

㪇㪅㪇㪈㪉㪐㫄㫄



㪇㪅㪇㪉

Figure 14. Experiment result of work posture rotation

㪪㪸㫄㫇㫃㪼㩷㪥㫆㪅㪌



㪇㪅㪇㪋



Repetition measurement accuracy We conducted the experiments in 50 times for each device and we can show the deviations in the measurements. The length of a baseline has a maximum of about 0.2mm variation. However, for example in Figure 12, the range of the deviation was 0.013mm in maximum.



㪇㪅㪇㪍

㪪㪈

㪿㽵

㪇㪅㪇㪍

㪈㪅㪌㪏㫦

㪄㪇㪅㪇㪉

Figure 11. Terminal lead measurement points

㪇㪅㪇㪎

㪈㪅㪇㪈㫦

㪈㪅㪊㪇㫦



㪇㪅㪇㪈㪎

㪿㽳

㪥㫆㪅㪋㪇㪋㪇

㪇㪅㪇㪈㪎

㪿㽴

㪿㽲㩷㩿㫄㫄㪀

㪇㪅㪇㪉㪌 㫄㫄

㪇㪅㪎㪋㫦

㪈㪅㪈㪉㫦

㱔䋲㪔㪈㪅㪌㪏㫦

㪇㪅㪇㪏

㪇㪅㪇㪉㪎 㫄㫄

㪇㪅㪊㪈㫦

㪊㪐

㪫㪼㫉㫄㫀㫅㪸㫃㩷㪥㫆㪅

Figure 13. Experiment result of work feed distance 452