P. S. Chaudhari, Prof. D. M. Patel, Dr. J. L.

Juneja / International Journal of Engineering Research

and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 4, July-August 2012, pp.1025-1028
Artificial Intelligence apply for prediction of Laser Cutting Process
– A review
P. S. Chaudhari1, Prof. D. M. Patel2, Dr. J. L. Juneja3
1
P. G. Student, U. V. Patel College of Engineering, Ganpat University, Kkerva, Mehsana,
2
Associate Professor, U. V. Patel College of Engineering, Ganpat University, Kkerva, Mehsana
3
Principal, Ahmedabad institute of technology, Ahmedabad - 382481

Abstract: There are many factors effective on performance of the laser cutting process. Identification of more
effective factors requires which will give the significant effect on the cutting quality of materials. In recent years the
researchers have explored the number of ways to improve the quality of cutting in the different lasers. This paper
reviews the research work carried out so far in the area of laser cutting process and artificial intelligence. In artificial
intelligence use for prediction model for given input data of laser cutting operation. Also artificial intelligence
beneficial for the time saving because no next experiments are required for measuring output parameters.

Key words: Artificial Intelligence, Neurons, Laser Cutting Processes.

1. Introduction: (2) The neodymium (Nd) is used for boring and

Laser: where high energy but low repetitions are
Laser is the acronym of Light Amplification required.
by Stimulated Emission of Radiation. Laser is light of (3) Neodymium Yttrium-Aluminum-Garnet
special properties, light is electromagnetic (EM) (Nd-YAG) laser is used where very high
wave in visible range. Main six processes are doing power is needed and for boring and
for Light Emission: Absorption, Spontaneous engraving. Both CO2 and Nd/ Nd-YAG
Emission, Stimulated Emission, Population lasers can be used for welding.
Inversion, Gain and Loss. Advantages of laser cutting over mechanical cutting
A laser device is consisted of: (1) Laser medium like include easier work holding and reduced
atoms, molecules, ions or semiconductor crystals; (2) contamination of work piece. Precision may be
Pumping process to excite these atoms into higher better, since the laser beam does not wear during the
quantum mechanical energy levels; and (3) suitable process. There is also a reduced chance of warping
optical feedback that allow the beam of radiation to the material that is being cut, as laser systems have a
either pass once through the laser medium or bounce small heat-affected zone. Some materials are also
back and forth repeatedly through the laser medium. very difficult or impossible to cut by more traditional
Lasers have now found applications in almost every methods. [2]
field of engineering, medicine, commercially etc. [1]
2. Artificial Intelligence:
Laser Cutting Process: Artificial Intelligence the branch of computer
Laser cutting is a technology that uses a science that is concerned with the automation of
laser to cut materials, and is typically used for intelligent behavior (Luger and Stubblefield, 1993) [3]
industrial manufacturing applications, but is also also the art of creating machines that perform
starting to be used by schools, small businesses and functions that require intelligence when performed by
hobbyists. Laser cutting works by directing the people (Kurzweil, 1990). [4] Artificial Intelligence is a
output of a high-power laser, by computer, at the branch of Science which deals with helping machines
material to be cut. The material then melts, burns, finds solutions to complex problems in a more
vaporizes away, or is blown away by a jet of gas, [1] human-like fashion. This generally involves
leaving an edge with a high-quality surface finish. borrowing characteristics from human intelligence,
Industrial laser cutters are used to cut flat-sheet and applying them as algorithms in a computer
material as well as structural and piping materials. friendly way. A more or less flexible or efficient
There are three main types of lasers used in laser approach can be taken depending on the requirements
cutting. established, which are use for prediction of result and
(1) The CO2 laser is suited for cutting, boring, also save the time and money of experiment work.
and Scribing engraving.
Applications: [5]

1025 | P a g e
P. S. Chaudhari, Prof. D. M. Patel, Dr. J. L. Juneja / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 4, July-August 2012, pp.1025-1028
 Control (air traffic), Dong-Gyu ANN, Gwang-won JUNG,
 Design (computer configuration), explained the influence of process parameter on
 Medical diagnosis, drilling characteristics of an AI 1050 sheet with a
 Instruction/training, thickness of 0.2 mm using a pulsed Nd: YAG laser
 Interpretation (speech), through numerical analyses (Software ABAQUS
 Monitoring (nuclear plant), V6.5) and experiments. For numerical analyses
 Planning (mission planning), parameters condition was Input voltage – 400v,
 Factory scheduling, duration time - 1.0 ms and 2.0 ms, frequency 4-12
 Prediction (weather), Hz, Nozzle diameter - 0.7 mm and Velocity of
 Repair (telephone). assisted gas - 345 m/s. Laser drilling experiments
were performed using a PC-NC controlled laser
Steps for Network Selection process for AI: [6] drilling system with investigate the variation of the
1. Prediction, diameter. The output parameters measured as the
2. Classification, diameter of the drilling hole, shape and the taper
3. Data Association, angle of the drill hole. All those parameters were
4. Data Conceptualization, measured via a Scanning Electron Microscope
5. Data Filtering. (SEM), comparison between numerical analyses and
experiment and find the optimal condition of drilling.
3. Literature Survey: From result, the optimal drilling condition was
Prof D. M. Patel, Dipesh Patel. Parametric determined at a plus duration time of 1 ms and a
Analysis of ytterbium: fiber laser cutting process. In pulse frequency of 12 Hz given the good result of
this report they mainly focus on cut quality and the drilling hole. [9]
cut quality mainly decided by surface roughness, kerf
width, and perpendicularity. The experiment was Nabil Ben Fredj et. al. generated ANN
carried out on 5mm thickness M.S. plate by varying model for prediction of surface roughness in grinding
the parameter like; laser power, gas-pressure, and process using table speed, depth of cut, grinding
cutting speed. The factorial design was used for wheel mesh size, dressing depth and number of
design of experiment and for find out the percentage passes as variables, with lowest calculated deviation
contribution of process parameter used Minitab 15 percentage between the predicted and the actual
software. Surface roughness was measured by values of the training data and the testing data, 2.47%
Surface roughness tester SJ-201 and kerf width was and 4.13% respectively. They stated that the
measured by equipment including digital camera and establishment of the regression models takes into
the UTHSCSA image tool version 3.0 program. [7] account the responses averages and therefore, can
give only an average model, which is not capable to
K. Abdel Ghany, M. Newishy, explained detect a particular tendency corresponding to specific
the cutting stainless steel by pulsed and CW Nd:YAG values or set of inputs. [10]
laser, it was shown that the laser cutting quality
depends mainly on the cutting speed, cutting mode, Ronny Pfeifer, Dirk Herzog, explained in
laser power and pulse frequency and focus position. this paper a suitable method to cut comparable thick
The cutting parameters that provided dross-free and 1mm Shape Memory Alloys sheets using an pulsed
sharp cut surface during pulsed laser mode were: Nd: YAG laser system with a high average power is
frequency = 200–250 Hz, duty cycle = 40%, actual presented. The impact of different laser and process
applied power = 210–250W (i.e., programmed peak parameters on the cut quality and the material
power, Pp = 880–1100W, cutting speed = 1–2 m/min, properties. A small kerf width (k = 150–300 µm) in
focus position = −0.5 to −1mm and nitrogen gas connection with a small angle of taper (θ < 2°) was
pressure = 9–11 bar. For CW mode, speed = 6–8 generated. Compared to short and ultra short-laser
m/min and power = 900–1100W with the same focus processing of SMA, high cutting speeds (v = 2–12
position and gas pressure. Using nitrogen gave mm/s) along with a sufficient cut quality (Rz = 10–30
brighter and smoother cut surface and smaller kerf µm) were achieved, offering the possibility to
than oxygen, although it is not economical. machine macro-SMA components (thickness: mm-
Increasing the frequency and cutting speed decreased range, cut length: cm/dm range) in an economic way.
the kerf width and the roughness of cut surface, while Compared to cw-laser machining a lower surface
increasing the power and gas pressure increased the roughness can be achieved. In comparison to short-
kerf width and roughness. Best cut quality was found and ultra short-laser processing of SMA the
at CW mode at speed from 6 to 8 m/min. Then, pulse drawbacks can be seen in the higher thermal impact
mode at low speed was 1–2 m/min. [8] of the laser–material processing on the SMA,

1026 | P a g e
P. S. Chaudhari, Prof. D. M. Patel, Dr. J. L. Juneja / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 4, July-August 2012, pp.1025-1028
resulting in a HAZ (dimension: 6–30 µm) which
again affects the material properties and the reduced
accuracy of the cutting process. [11]

Luke Huang et. al. developed a multiple

regression model for an on-line, real-time surface
roughness prediction system. They concluded that
with linear correlation coefficients of 0.940 and 0.933
for models MR16 and MR31, respectively, using the
experiment data, the predictor variables, such as feed
rate, vibration amplitude average, spindle speed, and
depth of cut, have strong linear correlation with the
predicted variable. Without the vibration data, the
prediction accuracy of the proposed multiple
regression model declines by about 1.55%.
Therefore, the use of the accelerometer is valuable.
The proposed regression model basically possesses Fig.: 2 Variation of kerf deviation and kerf width
accuracy of above 90% on predicting the in process with experiment number. [13]
surface roughness from feed rate, vibration amplitude
average, spindle speed, and depth of cut. [12]

4. Conclusion and Discussion:

Laser power and Cutting speed are plays an
important role in laser cutting. For cutting Mild Steel
by pulsed and CW Nd:YAG laser, it was shown that
the laser cutting quality depends mainly on the
cutting speed, cutting mode, laser power and pulse
frequency and focus position.

An Artificial Neural Network (ANN) for building a

feasible predicting model for Laser Cutting Qualities.
The Back-Propagation Neural Network with LM
algorithm has been applied to construct a
Mathematical Model wherein the laser-cutting
qualities are expressed as explicit nonlinear functions
of the laser control parameters. The GA is helpful to Fig.: 3 Effect of factor levels on multiple S/N
determine the best combination of laser-cutting ratio (0.9mm Al-alloy sheet). [13]
parameters for optimizing a specified cutting quality.
For applying AI we can use MAT LAB_2011 References
Software. [1] “Non- Conventional Machining Processes”
by P. K. Mishra, Published by Naroja
Publishing House. ( 3rd reprint 2005).
[2] http://en.wikipedia.org/wiki/Laser_cutting
[3] “Artificial intelligence: Structures and
strategies for complex problem solving” by
Luger and Stubblefield, published by
Benjamin/Cummings Pub. Co., Redwood
City, Calif. (2nd reprint:1993)
[4] “The age of intelligent machines” By
Kurzweil, published by MIT Press,
Cambridge, Mass. (Year- 1990)
[5] “Artificial Neural Network – Application”
by Chi Leung patrick Hui, Publisher-InTech,
Published – April 11,2011.
Fig.:1 LM16LP Model view with three layers.

1027 | P a g e
P. S. Chaudhari, Prof. D. M. Patel, Dr. J. L. Juneja / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 4, July-August 2012, pp.1025-1028
[6] James A. Freeman and David M. Skapura, [10] Nabil Ben Fredj, Ridha Amamou, Mohamed
Neural Networks, Algorithms, Applications Ali Rezgui, Surface Roughness Prediction
and Programming Techniques, Addison- Based Upon Experimental Design and
Wesley Publishing Company Inc., USA, pp. Neural Network Models, IEEE, SMC, TPIL2,
1- 4, 1991. 2002.
[7] Dipesh Patel, Prof. D.M.Patel. “Parametric [11] Ronny Pfeifer, Dirk Herzog, 2010, “Pulsed
Analysis of Ytterbium: fiber laser cutting Nd:YAG laser cutting of NiTi shape memory
process” - Ganpat university, kherva, alloys-Influence of process parameters”
Mehsana, Gujarat, India August-2010. Journal of Materials Processing Technology
[8] “Cutting of 1.2 mm thick austenitic stainless 210 (2010) 1918–1925.
steel sheet using pulsed and CW Nd:YAG [12] Luke Huang & Dr. Joseph C. Chen, “A
laser” by K. Abdel Ghany, M Newishy/2005, Multiple Regression Model to Predict In-
Journal of material processing and process Surface Roughness in Turning
technology, Vol-168, Page (438-447) 2005. Operation Via Accelerometer”, Journal of
[9] Dong-Gyu AHN, Gwang-won JUNG, 2009, Industrial Technology, Volume 17, No.- 2,
“Explained the influence of process 2001.
parameter on drilling characteristics of an AI [13] Avanish Kumar Dubey,Vinod Yadava,
1050 sheet with a thickness of 0.2 mm using “Optimization of kerf quality during pulsed
a pulsed Nd:YAG laser”, Transaction of laser cutting of aluminium alloy sheet”,
Nonferrous Met. Soc. China, s157-s163. journal of materials processing technology
204 ( 2 0 0 8 ) 412–418.

1028 | P a g e