Available online at www.sciencedirect.

com

ScienceDirect
Procedia Economics and Finance 15 (2014) 1563 – 1569

Emerging Markets Queries in Finance and Business

Productivity and Cycle Time Prediction Using Artificial Neural

Network
Cristian Gelmereanua, Liviu Morara, Stefan Bogdan*
a
Technical University of Cluj, Napoca, Cluj-Napoca, Romania

Abstract

The main purpose of the paper is to develop a neural network application that could predict the cycle time for High Speed
Machining. Increase productivity by increasing the speed of process has been imposed by the cutting progresses theory
and the fields related directly to the cutting process.
Thus, this procedure aims to an increasing productivity, lowering costs and execution time and also improving the quality
of parts.
© 2014 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
© 2013 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Emerging
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Markets
Selection and Queries
peer-review underinresponsibility
Finance and of Business local
the Emerging organization
Markets Queries in Finance and Business local organization

Keywords: productivity, Artificial Neural Network, cycle time prediction, High Speed Machining,

1. Introduction

It is a matter of common knowledge that higher productivity leads to a reduction in cost of production,
reduces the sales price of an item, expands markets, and enables the goods to compete effectively in the world
market. It yields more wages to the workers, shorter working hours and greater leisure time for the employees.

* Corresponding author. Cristian Gelmereanu

E-mail address: emicampean@gmail.com.

2212-5671 © 2014 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of the Emerging Markets Queries in Finance and Business local organization
doi:10.1016/S2212-5671(14)00626-1
1564 Cristian Gelmereanu et al. / Procedia Economics and Finance 15 (2014) 1563 – 1569

In fact the strength of a country, prosperity of its economy, standard of living of the people and the wealth of
the nation are very largely determined by the extent and measure of its production and productivity.
Productivity can be defined in many ways. Some of them are as follows:
• Productivity is nothing but the reduction in wastage of resources such as labor, machines, materials,
power, space, time, capital, etc.
• Productivity can also be defined as human endeavor (effort) to produce more and more with less and less
inputs of resources so that the products can be purchased by a large number of people at affordable price.
• Productivity implies development of an attitude of mind and a constant urge to find better, cheaper, easier,
quicker, and safer means of doing a job, manufacturing a product and providing
service.
• Productivity aims at the maximum utilization of resources for yielding as many goods and services as
possible, of the kinds most wanted by consumers at lowest possible cost.
• Productivity processes more efficient works involving less fatigue to workers due to improvements in the
layout of plant and work, better working conditions and simplification of work. In a wider sense productivity
may be taken to constitute the ratio of all available goods and services to the potential resources of the group
(Salomon, C.).

2. Ways to increase productivity using the cutting parameters

One way to increase productivity is by increasing the working regime parameters. Among these may be
mentioned cutting depth (t), the feed (s) and the main cutting speed (v) measuring the volume "V" of material
that will be removed per unit of time. In order to decrease the cycle time we have to take into account the
quality of the surface, the required accuracy, the tool durability, the workpiece dimensions, the removed chips
size etc. One of the possibilities to increase the profitability of the process is by raising cutting speed. This has
always been a constant concern of cutting process specialists. This paper deals with this topic, highlight the
main advantages of this method and trying to generate the cycle time using also neural networks. It was
designed a stand in the Engineering Design and Robotics Laboratory from the Technical University of Cluj
Napoca (Figure 1.) to conduct research regarding the productivity. We want to implement an algorithm that
leads to an increase cutting regime having as signal the vibration level.

Fig.1. Milling machine stand

 Cristian Gelmereanu et al. / Procedia Economics and Finance 15 (2014) 1563 – 1569 1565

3. Literature review

The first concept of High Speed Machining (1931) was presented by Carl J Salomon (Salomon, C.,1931),
who is considered as the father of high speed machining. The results of his experiments, carried out on non-
ferrous metals such as Aluminium, Copper and Bronze, using helical milling cutters at speeds up to 18067
m/min (surface metres per minute) revealed that with the increase of cutting speed, the cutting temperature
increases up to a maximum value close to the melting point of the material and then decreases with further
increase in speed (King, R.I, 1984). Bhattacharyya et.al investigated on cutting force-based real-time
estimation of tool wear in face milling using a combination of signal processing techniques (Bhattacharyya
et.al, 2007). The tool wear in high speed milling using detection process approach has been done by Kious
et.al. Estimation of tool wear during CNC milling using neural network-based sensor fusion was implemented
by Ghosh et.al. On-line metal cutting tool condition monitoring using multi-layer perceptron neural networks
was studied by Lister and Dimla. Tool condition monitoring using artificial intelligence methods has been
carried out by Balazinski et.al. Cho and Ko estimated tool wear length in finish milling using a fuzzy
inference algorithm. Intelligent process supervision for predicting tool wear in machining processes was done
Alique by and Haber. Ning and Veldhuis analyzed mechanistic modeling of ball end milling including tool
wear. (W. Yan, 1999) Development and processes manufacturing based on cutting speed from 10 m/min in
1800, up to 100 m / min in 1900, a thousand feet per minute in 1980 to several thousand in 1994 (figure 2)
imposed default also a tendency to continue studying these processes in terms of the elements that contribute
directly to their achievement, their deployment conditions, the level of cutting regime.

Fig.2. High Speed Machining

A very general definition of high speed processing method is proposed by French researcher Alain Defretin
in his book “Usinage a Grande Vitesse”. (Ivan, R., 2011) "Processing of high speed milling includes all
processes (high feed rates or high speed spindle rotation), processes that are superior to those that are
considered today controlled from a practical standpoint."( Defretin, A.) In the case of high-speed processing,
the main axis of the spindle is greater than 15,000 revolutions per minute, as illustrated above.
Growth rate has a reverse effect as it reduces cutting forces, it improves the quality of machined surfaces
and improves the natural fragmentation of chips, etc.. In this sense, today more than yesterday and less than
tomorrow, high speed cutting technology is recognized as one of the key processing technologies in order to
1566 Cristian Gelmereanu et al. / Procedia Economics and Finance 15 (2014) 1563 – 1569

obtain increased productivity and a precision manufacturing performance. In the case of high-speed
processing, the processing speed is higher than the propagation speed of the heat in the work piece. Below
there is a comparison of speeds used during machining some selected materials using conventional and HSM
methods.

Table 1. Conventional vs. High Speed Machining (http://fstroj.utc.sk/journal/engl/papers/034_2002.pdf)

Solid Tools (end mills, drills), WC, Indexable Tools (shell mills, face
coated WC, PCD, ceramic mills) WC, ceramic, CBN, PCD
Work material Typical cutting High cutting Typical cutting High cutting
Speed [m/min] speed [m/min] speed [m/min] speed [m/min]
Aluminium >305 (WC, >3050 (WC, >610 >3658 (WC,
PCD) PCD) PCD)
1219 (sialon,
soft 152 366 366
Cast iron ceramic)
ductile 107 244 244 914 (ceramic)
Free mach steel 107 366 366 610
alloy 76 244 213 366
Steel stainless 107 152 152 274
Hardness 30 (WC) 91 46 (WC), 183
24 122
HRC65 (CBN, ceramic) (CBN, ceramic)
Titanium 38 61 46 91
Superalloy 46 76 84 (WC) 213 366 (sialon,
(sialon) ceramic)
In fact, the question is what are the advantages of high speed milling compared to conventional milling
(table 2)

Table 2. The advantages/ disadvantages of using high-speed milling

Advantages of using high-speed milling Disadvantages of using high-speed milling

Economic efficiency and increased production
A high-quality machined surface with a roughness
of up to 0.2 m;
Possible to obtain rapid milling of thin walls up to
0.02 mm thick to 20 mm;
Temperature transferred into the surface layer of
the workpiece is less than the feed rate; the heat
doesn’t have time to propagate in the part;
Achieve higher dimensional precision than with
conventional process;
The need for qualified personnel. It required a good
knowledge of the cutting process, of the equipment;
Acceleration and deceleration are very fast processes,
many stops and starts that lead to premature wear of
the machine tools;
The consequences are more severe for human errors
or software failures;
Need to take additional safety measures (doors closed
during processing, tools verification etc.)

Thus, this procedure aims to increase productivity, lower the costs and also improve the quality of the
made parts (table 3). (http://www.slideshare.net/hurco/high-speed-machining-hurco-imts-2012)
 Cristian Gelmereanu et al. / Procedia Economics and Finance 15 (2014) 1563 – 1569 1567

Table 3. Traditional Processing vs High Speed Machining

Traditional Processing High Speed Machining High Speed Machining

(12K) (12K) (18K)
Hourly Shop Rate 60.00$ 60.00$ 60.00$
Hourly Burden Rate 35.00$ 35.00$ 35.00$
Cycle Time 11min 16sec 2min 51sec 1min 54sec
Cost to Produce 6.57$ 1.67$ 1.11$
Quoted Price 13.50$ 13.50$ 13.50$
Profit per Part 6.93$ 11.83$ 12.39$
Profitx100 pieces 693.00$ 1183.00$ 1239.00$
Profitx500pieces 3465.00$ 5915.00$ 6195.00$

However, this quality is directly influenced by the state of the cutting tool, which is exposed to wear during
machining. Thus control of wear during the cutting process proves to be important. What unites all these
terms "target" are the cutting parameters, which can optimize the milling operation (reduced wear, reduced
dynamic aspects) and thus reduce manufacturing costs by placing the operation in a stable and intermediate
processing zone, and reducing the cycle time, while increasing productivity.
For this are consider the following parameters: (Ivan, R., 2011)
• Spindle speed (rpm)
• Feed rate (mm/min)
• Depth of cut (mm)
• Cycle time (min)
The experiments were conducted in a ZXX6350ZA vertical axis milling machine using tool life 20 mm
diameter cutter mill with 4 cutter inserts. The cutter mill was made by high speed steel-E (HSS-E). Work-
pieces used for this experiment was Al 7075 (Table 4) (http://fstroj.utc.sk/journal/engl/papers/034_2002.pdf).

Table 4. Cutting data

No. Spindle speed (rpm) Feed Rate Depth of cut (mm) Cycle time (min)
(mm/min)
1 95 132 0.2 26.4
2 360 98 0.6 45.5
3 565 200 0.2 38.7
4 950 98 1 15.7
5 1500 132 0.4 14
6 18000 143 1.5 15.6
7 24100 1360 1 35
8 24050 1390 0.4 5
9 35000 1750 0.2 8.02
10 40000 2120 3 12
11 40500 2200 1.5 3.5
With the parameters in Table 4 was tried modeling and generation of the cycle time using ANN reducing
cycle time due to use of high-speed milling.
1568 Cristian Gelmereanu et al. / Procedia Economics and Finance 15 (2014) 1563 – 1569

4. Tool life determination using Artificial Neural Networks (ANN)

An Artificial Neural Network (ANN) is a mathematical model or computational model that is inspired by
the structure and/or functional aspects of biological neural networks. According to Simon Haykin, neural
network is “a massively parallel distributed processor with the natural tendency to store experimental
knowledge and make them available for use”. In most cases an ANN is an adaptive system that changes its
structure based on external or internal information that flows through the network during the learning phase
(Haykin, 2009). The similarity with the human brain consists in learning request before acquiring knowledge.
Learning takes place through the entries that are brought in the system. Knowledge is stored in connections
interneurons, ie neural weights, which stores the results (Beale et al., 2011). We used the data from table 4,
having in the input data the spindle speed (rpm), the feed rate (mm/min)and the depth of cut (mm), and in the
output layer the cycle time. The Levenberg-Marquardt algorithm was used for training and the MLP
(Multilayer perceptrons) training was done automatically using the elaborated software package.(Campean,
E., 2012) The ANN consists in 5 neurons. simulating first with 11 pairs of input in the input layer. It is noted
that the number of neurons chosen, the number of intermediate layers and the activation functions were well
chosen because, after only 3 iterations (epochs), neural network has reached a result in the indicated error,
having a running time of five seconds. It can be seen that the error of 10-2 is satisfactory. The error gradient is
10-15. Along with this result, is made available a graphic illustration of the training process. Table 5 show the
result:

Table 5. The ANN results

No. Spindle Feed Rate Depth of cut Cycle time ANN Error
speed (rpm) (mm/min) (mm) (min) cycle time (%)
1 95 132 0.2 26.4 25.71 0.0069
2 565 200 0.2 38.7 41.01 -0.0231
3 950 98 1 15.7 15.96 -0.0026
4 24100 1360 1 35 35.62 -0.0062
5 24050 1390 0.4 5 5.31 -0.0031
6 35000 1750 0.2 8.02 8.36 -0.0034
7 40000 2120 3 12 11.91 0.0009
8 40500 2200 1.5 3.5 2.92 0.0058
9 20000 900 1.5 8.3 7.52 0.0078
10 10095 4040 1 14.7 12.56 0.0214
11 8514 5110 1 12.3 12.01 0.0029
12 12038 4820 0.4 3.6 3.02 0.0058
13 13403 5360 0.4 3.5 3.02 0.0048
14 11196 3360 1.5 4.2 3.5 0.007

The results shown in table 5 indicate approximately the same values, what means that the ANN approach
is suitable for such application. The great advantage of the ANN is the time needed to obtain the results.
 Cristian Gelmereanu et al. / Procedia Economics and Finance 15 (2014) 1563 – 1569 1569

5. Conclusion

In this study, ANN for modeling and predicting cycle time for milling parts made of Aluminum material
was developed. Given the accuracy that was achieved it is safe to conclude that the ANN approach is suitable
for such application. It was found that (ANN) prediction correlates very well with the experimental results.
Given the results obtained by the ANN simulation was designed an algorithm that involves increasing the
cutting regime during machining. Maximum cutting regime parameters (cutting depth for beginning and
advanced) will be equal to the acceptable level of vibration.

Acknowledgements

This paper was supported by the project "Improvement of the doctoral studies quality in engineering
science for development of the knowledge based society-QDOC” contract no. POSDRU/107/1.5/S/78534,
project co-funded by the European Social Fund through the Sectorial Operational Program Human Resources
2007-2013.

References
Salomon, C., Critical Machining Velocities. 1931: German Patent No. 523594.
King, R.I. and R.L. Vaughn. A synoptic review of High-speed machining from Salomon to the present. in High speed machining :
presented at the winter annual meeting of the American Society of Mechanical Engineers 1984. New Orleans, Louisiana: ASME
(345 E. 47th St., New York 10017): p. 1-13.
P. Bhattacharyya, D. Sengupta, S. Mukhopadhyay, “Cutting forcebased real-time estimation of tool wear in face milling using a
combination of signal processing techniques,” Mechanical Systems and Signal Processing, Vol 21, 2007, pp 2665–2683.
M. Kious, A. Ouahabi, M. Boudraa, R. Serra, A. Cheknane, “Detection process approach of tool wear in high speed milling,”
Measurement, Vol 43, 2010, pp 1439–1446.
N. Ghosh, Y.B. Ravi, A. Patra, S. Mukhopadhyay, S. Paul, A.R. Mohanty, A.B. Chattopadhya, “Estimation of tool wear during CNC
milling using neural network-based sensor fusion,” Mechanical Systems and Signal Processing, Vol 21, 2007, pp 466–479.
D.E. Dimla Sr. a, P.M. Lister, “On-line metal cutting tool condition monitoring.II: tool-state classification using multi-layer perceptron
neural networks,” International Journal of Machine Tools & Manufacture, Vol 40, 2000, pp 769–781.
M. Balazinski, E. Czogala, K. Jemielniak, J. Leski, “Tool condition monitoring using artificial intelligence methods,” Engineering
Applications of Artificial Intelligence, Vol 15 , 2002, pp 73–80.
T. J. Ko, D. W. Cho, “Estimation of tool wear length in finish milling using a fuzzy inference algorithm,” Wear, Vol 169, 1993, pp 97-
106.
R. E. Haber, A. Alique, “Intelligent process supervision for predicting tool wear in machining processes,” Mechatronics, Vol 13, 2003, pp
825–849.
W. Yan, Y.S. Wong, K.S. Lee, T. Ning, “An investigation of indices based on milling force for tool wear in milling,” Journal of Materials
Processing Technology, Vol 89, 1999, pp, 245–253.
Ing. IVAN Radu Alexandru, „ Studiul comportamentului dinamic al sistemului port-scula-scula aschietoare-piesa in cazul operatiei de
frezare cu viteze inalte”, Teza de doctorat, Universitatea Transilvania din Brasov, Brasov, 2011
Defretin, A., Usinage a grande vitesse, available at http:/www.techniques-ingenieur.fr/dossier/usinage a grande vitesse/BM7180.
http://fstroj.utc.sk/journal/engl/papers/034_2002.pdf
http://www.slideshare.net/hurco/high-speed-machining-hurco-imts-2012
Campean E., Morar L., Baru P., Bordea C., Pop D., Aspects regarding some simulation models for logistic management, Procedia
Economics and Finance 00 (2012) 000–000, EFQM, Targu Mures, Romania
E. Ciupan, A Model for the Management of a Supply Activity, Based on Statistical Data, Quality and Innovation in Engineering and
Management, Cluj Napoca, 2011, p.249-253.
Russell S., Norvig, P.,” Artificial Intelligence: A Modern Approach”. Prentice Hall, 2002.
Simon Haykin, “Neural networks and learning machines”, Volume 10, Prentice Hall, 2009
Mark Hudson Beale, Martin T. Hagan, Howard B. Demuth, Neural Network Toolbox™ User’s Guide