Professional Documents
Culture Documents
com
Available
Available onlineonline at www.sciencedirect.com
atonline
www.sciencedirect.com
Available online
Available at www.sciencedirect.com
at www.sciencedirect.com
ScienceDirect
ScienceDirect
Procedia CIRP 00 (2018) 000–000 www.elsevier.com/locate/procedia
Procedia
Procedia CIRP CIRP 00 (2018)
00 (2017)
Procedia CIRP 000–000
000–000
77 (2018) 155–158 www.elsevier.com/locate/procedia
Procedia CIRP 00 (2018) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
[11]. As their main specifications show (see Table 2) the Canon 3. Case study A: Shear angle estimation
lens ensures better magnification (up to 5:1) and closer focus-
ing distance, although, according to our estimations and tests, In this case study, the predictions of the theoretical shear an-
the depth of field (DOF) is significantly shallower. gle models are compared with the estimated shear angles based
on high-speed camera recordings. The shear angle has signif-
Table 1. Specification of Photron Fastcam SA5 high-speed camera [9] icant role in the theoretical cutting force expressions, thus it
Frame rate 50 − 775,000 fps also affects the dynamics of metal cutting and the occurrence of
Max. resolution 1024 × 1024 pixel (up to 7000 fps) machine tool chatter [1–4].
Shutter 1µs The chip formation process and the shear angle variation
Memory 32 GB (recording duration: 21 − 3 s) can be investigated during stationary cutting with constant chip
thickness (h), which can be realized by orthogonal planning
process. In orthogonal cutting, the material in front of the tool
Table 2. Comparison of the applied macro lenses [10,11] is sheared over a primary zone that is approximated by the shear
plane at the angle Φ. In other words, the shear angle determines
Nikon Nikkor 200 mm Canon MP-E 65 mm
the magnitude and the direction of the cutting force F acting on
Max. Aperture f/4 f/2.8 the tool [12,13].
Magnification 1:1 1:1−5:1
Estimated Depth of Field ≈ 0.55 mm ≈ 0.4 − 0.05 mm
Closest focusing distance 0.5 m 0.24 m a) b)
primary αr
zone
In order to provide sufficient lighting, which is crucial in shear
vc plane
high-speed camera measurements, two Vision Device VD7000 tool tool
LED (max. power 72W) lamps and an endoscope-like Hayashi
h
HDF7010 Fiber Optic LED (max. power 155W) system were Φ Fx
applied. Note, that by increasing the frame rate, the brightness
workpiece Fz
of the image decreases significantly. This effect is illustrated
in Fig. 1/a, where average brightness (Baverage ) obtained from
the HSB values of a test image at different frame rates is pre-
sented for both lenses. Additionally, due to the limited memory Fig. 2. (a) Merchant’s model of cutting forces (b) measured cutting forces ac-
tion on the tool
capacity, the resolution also decreases (see Fig. 1/b). Hence,
the applied frame rate should be selected carefully considering
In the model of Merchant (see Fig. 2/a), the component Fs
also the cutting speed and/or the spindle speed. For instance,
of the cutting force F aligned with the shear plane is obtained
we applied frame rate of 7,000−12,000 fps for chip formation
from the yield shear stress of the material and the shear plane as
(up to vc = 10,000 mm/min) and 10,000−30,000 fps for chatter
Fs = τ s A s . This yields that the other components of the cutting
detection with Ω = 6,000 − 15,000 rpm (see Fig. 1).
force can also be expressed using the tool geometry, namely the
rake angle αr and the average friction angle βa [12,13].
chip a)
formation 3.1. Shear angle models
chatter detection
In the literature, the two most commonly applied shear an-
gle models are based on the maximum shear stress principle
Canon MP-E 65 mm macro (MSSP) and the minimum energy principle (MEP). The MSSP
assumes that the shear occurs in the direction of the maximal
Nikon Nikkor 200 mm macro shear stress, which has an angle of π/4 with the cutting force.
According to Merchant’s circle in Fig. 2, the shear angle can be
expressed using the tool geometry as
π
b) Φ MS S P = − (βa − αr ). (1)
4
However, the MEP states that the power of the cutting force
P = Ft vc should be minimal. This yields, that the shear angle
chip becomes
formation
π β a − αr
chatter detection Φ MEP = − . (2)
4 2
Fig. 1. The dependency of a) average brightness of a test image and b) maximal The planning tests were performed on a single aluminum
resolution on applied frame rate and the applied frame rate regimes (A2024-T351) rib of thickness w = 2 mm using the feed motion
of a NCT EmR-610Ms CNC milling machine. The workpiece
Szabolcs Berezvai et al. / Procedia CIRP 77 (2018) 155–158 157
S. Berezvai et al. / Procedia CIRP 00 (2018) 000–000 3
tool High-speed Fig. 4. Comparison of the theoretical shear angle models to the image process-
camera system ing method at different chip thicknesses (w = 2 mm, αr = 15◦ , α f = 10◦ ,
w h0 vc = 10,000 mm/min) during orthogonal cutting
z
y
x
vc 4. Case study B: Chatter detection
Canon MP-E
workpiece 65 mm f/2.8 In this case study, a chatter detection method is proposed us-
dynamometer ing high-speed camera recordings. The detection of machine
tool chatter is an actual problem in the industry, for which sev-
Fig. 3. The schematics of the experimental setup eral chatter identification systems are proposed in the literature,
e.g. using microphones and dynamometers [14,15]. Alterna-
tively, machine tool chatter can also be detected based on high-
Table 3. Parameters of the orthogonal cutting speed camera recordings for which a video-based image detec-
Tool Carbide tool, tion algorithm was developed.
wtool = 5 mm, αr = 15◦ , α f = 10◦
Workpiece Aluminum A2024-T351 4.1. Method of chatter detection
w = 2 mm
Process Orthogonal planning, vc = 10,000 mm/min
h = 0.05, 0.1, 0.15, 0.2, and 0.25 mm The video-based chatter detection method was applied dur-
High-speed camera 10,000 fps, Canon MP-E 65 mm macro lens ing stable and unstable milling operations, which were carried
magnification: 5:1 out on an NCT EmR-610Ms CNC milling machine with a ZCC
EMP01-016-G16-AP11-02 tool. The operation parameters cor-
responding to the stable and unstable milling tests are listed in
The chip formation was investigated using high-speed cam- Table 4. The milling tests were also recorded with high-speed
era recordings with 7,000 fps. Since the size of the investi- camera at 10,000 fps using Nikon 200 mm f/4D macro lens
gated phenomena was small, the camera was equipped with a to ensure adequate magnification and depth of field. It should
Canon MP-E 65 mm macro lens to ensure maximal magnifica- be emphasized that reflection from the tool surface is required
tion (5:1). The shear angle was estimated using a displacement- for the image processing method, therefore the lighting system
based image processing technique, where the displacement field should be placed appropriately.
was calculated from two subsequent frames. After applying
Wiener-filtering, the occurrence of the vertical displacement Table 4. Parameters of the milling tests
values indicated the points of the shear plane. The shear angle
Tool ZCC EMP01-016-G16-AP11-02
was obtained from the straight line fitted to the detected points N = 2, d = 16 mm
of the shear plane [5]. Workpiece Aluminum A2024-T351
Process Down-milling, Ω = 8000 rpm, fz = 0.05 mm/tooth
Stable milling: ae = 8 mm, a p = 1 mm
3.3. Measurement results
Unstable milling: ae = 12 mm, a p = 2 mm
High-speed camera 10,000 fps, Nikon 200 mm f/4D macro lens
The detected shear angle and the corresponding standard de- length of recording: 0,1925 s, 1925 frames
viation values can be compared to the prediction of the theoret-
ical models presented in Eqs.(1)-(2). The results show (see Fig.
4), that for small chip thicknesses the prediction of the MEP Based on the high-speed camera recordings, machine tool
model and the image processing method show perfect corre- chatter was detected using the following algorithm. After crop-
spondance. However, during segmentation at high chip thick- ping a small part of the video with well-detectable light pattern
nesses (see the gray shaded area in Fig. 4) both the qualitative on the tool, the video was binarized to black-and-white with
and quantitative prediction of the models became inaccurate. threshold of 0.7 (see Fig. 2). The position of the centre of mass
It should be noted that flying chips and built-up edges caused C(xc , yc ) was calculated for each frame from which the relative
inaccuracy in the image detection algorithm. This may be the displacement of the ith frame in x-direction could be obtained
reason of the high standard deviation values obtained in case of as ∆xc [i] = xc [i] − xc [1]. The spectrum of the vibrations was
image detection on the high-speed camera recordings. obtained using FFT.
158 Szabolcs Berezvai et al. / Procedia CIRP 77 (2018) 155–158
4 S. Berezvai et al. / Procedia CIRP 00 (2018) 000–000
5. Conclusions
raw video
Acknowledgements