Anda di halaman 1dari 10

Proceedings of the ASME 2015 International Design Engineering Technical Conferences &

Computers and Information in Engineering Conference

August 2-5, 2015, Boston, Massachusetts, USA




Sushrut Pavanaskar Sara McMains∗

Department of Mechanical Engineering Department of Mechanical Engineering
University of California, Berkeley University of California, Berkeley
Berkeley, California 94720 Berkeley, California 94720
Email: Email:

ABSTRACT relatively little formal study [1]. We lack tools, standards, and
This paper describes our work on analyzing and modeling metrics to make tangible progress in saving energy to comply
energy consumption in CNC machining with an emphasis on with regulations [2].
the geometric aspects of toolpaths. We address effects of
geometric and other aspects of toolpaths on energy consumed in In this work, we focus on energy use in the CNC milling
machining by providing an advanced energy consumption model process. We have observed that commercial CAM software
for CNC machining. We performed several controlled machining for generating CNC milling process plans (toolpaths, machining
experiments to isolate, identify, and analyze the effects of various parameters, etc.) does not consider energy consumed to machine
aspects of toolpaths (such as path parameters, angular change, a component as an objective beyond conventional objectives
etc.) on energy consumption. Based on our analyses, we such as cycle time. Nor do commercial toolpath planning
developed an analytical energy consumption model for CNC software tools provide energy consumption estimates for the
machining that, along with the commonly used input of material various alternative toolpath strategies they offer when planning
removal rate (MRR), incorporates the effects of geometric toolpaths. The potential impact of such data is substantial,
toolpath parameters as well as effects of machine construction especially in light of the renewed interest in CNC machining,
when estimating energy requirements for a toolpath. We also as users well beyond the traditional automotive and aerospace
developed a simple web-based software interface to our model, industries embrace the process on a massive scale [3].
that, once customized for a particular CNC machine, provides
energy requirement estimates for a toolpath given its G/M code.
Such feedback can help process planners and CNC machine The paper is organized in six sections, starting with a
operators make informed choices when generating/selecting brief description of prior work. We then propose an energy
toolpath alternatives using commercial CAM software. consumption model for CNC milling that addresses some of the
needs identified above by explicitly considering and analyzing
the geometry of the toolpath followed in the process. We
INTRODUCTION also describe the design and the user interface of our energy
Although improving the productivity of CNC milling has analyzer software that can be used in process planning to quickly
been extensively researched, energy consumption and other estimate energy consumption for a given toolpath (G/M code)
environmental impacts of manufacturing processes have received for the target CNC machine. Finally, we provide three test cases
that highlight potential energy savings that may currently be
overlooked in routine process planning.
∗ Address all correspondence to this author.

1 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

We use the following nomenclature in this paper: Energy consumed by a CNC machine (Emachine ) can be
Ns : Spindle speed divided into:
Fi : “i” axis component of feedrate
d : Depth of cut 1. Econst : components that consume energy at a constant rate,
w : Width of cut irrespective of the cutting operation (e.g. lights, computer,
D : Cutter diameter fans), and
Fc : Cutting force 2. Evariable : components that consume energy at a variable rate,
Pi : Instantaneous power consumed by component i depending on the cutting operation (e.g. energy for material
Ei : Energy consumed by component i removal, table movements, spindle rotation).
Ti : Time consumed by component i
MRR : Material removal rate in cubic inches per second. Thus, we write:

Emachine = Econst + Evariable . (2)

Gutowski et al.’s seminal energy consumption model for We propose to compute the total energy requirement of a
CNC milling, tying energy consumption to material removal rate machining operation as a sum of all machining (e.g. cutting) and
(MRR) [4], has since been expanded upon by many researchers. non-machining (e.g. ramping the axes) activities in the operation.
These MRR-based models compute either the specific energy If the machining activity is represented by G/M code (NC code),
(i.e. energy required to remove a unit volume of material) we thus compute the energy requirement as a summation of
in CNC milling [5, 6, 7], or estimate energy requirements for energy requirements of all lines (blocks) of the code:
machining an entire component [8, 9]. A detailed survey of these
models can be found in [9].
Emachine = ΣEblock .
A generalized representation of most existing models can be
made in the form of:
For each block of NC code, we write the energy consumption
equation as:
E = ∑ PowerStatei · timei , (1)
Eblock = Pblock · tblock , (3)
where E is the total energy required, and PowerStatei and timei
are various “states” of the machine that draw different amounts where Pblock is the electric power drawn by the machine from the
of power from the grid depending on the sub-systems in action grid and tblock is the machining time for that block (activity) in
for the corresponding time intervals. the NC code. We compute tblock using an acceleration value of
Most analytical process models cited above consider the a = 1m/s2 , which is the maximum acceleration of most CNC
actual metal cutting process as an aggregate function of MRR but machines used today. We further divide Pblock into:
rarely isolate the contribution of individual process parameters
(Newman et al. proposed this as a promising area to investigate
[7]). As observed by [10,11], physical machine construction and Pblock = Pconstant + Pvariable , (4)
toolpath geometry also affect energy consumed in machining,
but most energy models consider only simple toolpaths, thus similar to eq. 2, as follows:
ignoring the effect of toolpath parameters, toolpath strategy,
Pconstant constitutes all components in the machine that draw
milling direction, or machine construction when modeling
power at a constant rate, irrespective of the cutting operation or
energy consumption.
toolpath. Examples of such components include the controller,
We propose an energy consumption model for 3-axis CNC
lights, coolant pump, etc. For a block of NC code, depending on
milling machines in the next section that builds on some of these
the state of the machine, one or many of these contribute a fixed
prior models while additionally incorporating some relations
amount of power to the total power consumption by the machine.
derived in Altintas’s work on a cutting-force model for predicting
In general,
tool-breakage [10]. With our enhanced energy consumption
model, we aim to perform machine-specific, individual process
parameter based, geometric analysis of NC toolpaths. Pconstant = Ptare + Pcoolant_pump . (5)

2 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

Pvariable constitutes the power drawn by the machine at a Thus,
variable rate, according to the activity in the block. Pvariable
is comprised of two major components: Pmoving_axes , the power PCoulomb_ f riction , Pdamping , Pmaterial_removal ∝ Fc , and
drawn by the machine merely to move (table movement, spindle
axis movement, and spindle rotation), and Pcutting_ f orce , the Fc ∝ MRR. (10)
additional power drawn from the grid to exert the cutting force Thus, Pcutting_ f orce ∝ MRR.
for removing material (if cutting happens). So,
These inferences of linear proportionality were also verified
Pvariable = Pmoving_axes + Pcutting_ f orce . (6) experimentally in [10, 13, 11].
In order to use our model effectively for a range of CNC
machines with possibly different scales and configurations, we
Hence, consolidating the above equations, incorporate several experimentally-determined machine-specific
constants into our model. These constants, called customization
Pblock = Ptare + Pcoolant_pump + Pmoving_axes + Pcutting_ f orce . (7) constants, are introduced for all proportional relations derived
above in Equations 8 and 10. By introducing customization
constants, we obtain linear equalities between the process
We determined from our machining experiments that the parameters and their corresponding power usage contributions.
energy required for mere axis movements (starting from rest) is For each such relationship, e.g. between any two variables ζ and
directly proportional to the programmed feedrate F. Similarly, ψ that are linearly proportional to each other, we introduce two
the energy required to rotate the spindle is proportional to the constants KA and KB such that ζ = KA + KB · ψ. This form
programmed spindle speed Ns , as seen in our experiments [12] of linearity is similar to that used in models by [14] and [6].
and as reported in [9]. Thus, Also, we compute the customization constants and the quantities
Pmoving_axes and Pcutting_ f orce as a summation of those for each of
the three principal axes (X, Y, and Z). Such an individualized
Pmoving_axes = Pspindle_rotation + Paxes_movement , calculation for each axis helps in accurate consideration of
where, toolpath geometry, since each axis may have a different mass
(8) and hence a different value for customization constants.
Pspindle_rotation ∝ Ns ,
Paxes_movement ∝ F. Using the customization constants, our final model for Pblock
can be written as the equation:

Pcutting_ f orce , the additional power drawn from the grid Pblock = Ptare + Pcoolant_pump + (KAmove_i + (KBmove_i · Fi ))
to exert the required cutting force Fc , can be computed as

follows. From first principles, not all the drawn power (energy)
+(KAspindle + (KBspindle · Ns )) + ∑ (KAcut_i + (KBcut_i · MRR)).
is converted into cutting force; some energy is always lost in i=x,y,z
overcoming Coulomb friction, in damping, and in waste heat. (11)
This treatment is identical to Eq. 1 in [10], where the authors
state that an axis motor’s useful torque is spent in overcoming
inertia, damping, friction, and the rest is converted to cutting Use of customization constants enables us to use the same
torque. We have considered the inertia component in computing underlying model in Equation 11 for any CNC machine. Further,
Pmoving_axes and thus it is not included again in computing since the constants are determined experimentally, the model
Pcutting_ f orce . Thus, excluding the unmeasured waste heat, we does not rely on any manufacturer specifications or machine-tool
write: handbooks. Also, a machine can be customized periodically to
ensure that the model predicts energy estimates for the current
state of the machine. Since the customization constants for
Pcutting_ f orce = PCoulomb_ f riction + Pdamping + Pcutting . (9) KAcut_i and KBcut_i are dependent on the physical properties of
the materials (cutter and workpiece), they are specific to each
We note from [13] that the cutting force required to cut the tool/workpiece material combination.
material is directly proportional to the material removal rate In the next section, we explain the various design decisions
MRR of the cutting process. MRR is a function of width of cut w, we made when developing the energy analyzer software based on
depth of cut d, and feedrate F. From [10], Coloumb friction and this energy model to make machine-specific energy consumption
damping are directly proportional to the exerted cutting force Fc . predictions.

3 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

ENERGY ANALYZER SOFTWARE analyzer as a web-based service to allow its use in a web
In the previous section, we developed an analytical energy browser. Thus, it can be readily accessed and used on
consumption model for 3-axis CNC milling that predicts the a shop floor, in geographically distributed manufacturing
power drawn by a CNC machine at any stage of operation. Here, sites, places with limited computational power, etc. All data
we use this model to develop a software tool that analyzes a processing happens on a remote server and only the resulting
planned machining operation to estimate the energy requirement plots are rendered on the client machine.
of the entire operation. With these estimates, we aim to
provide a means to correlate parameter and toolpath choices
made in process planning of an operation with their effects on
the energy (electricity) required for the operation. The intended
users of this software are process planners, who can use this
software along with toolpath planning software to make process
planning decisions by simultaneously considering environmental
and productivity objectives. We first briefly describe the various
software design decisions made when developing this software.

Software Design
We made the following decisions when developing the
energy analyzer software such that it can be readily used during
process planning on a shop floor.
1. Individualized analysis: we designed this software such
that the same underlying analytical model is used to make
machine and case-specific energy predictions for a range
of CNC machines. Therefore, in addition to the use
of machine specific customization constants in the energy
model, the software accepts and incorporates case-specific
process parameters such as typical percentage overlap, tool
diameters, etc.
2. NC code (G/M commands) as input (as opposed to the part
geometry): This software is intended to be used after a
process planner has generated an NC code for a component
using any CAM software. Instead of computing “typical”
energy required for manufacturing a part, based only on
the part geometry as done in much existing CAD/CAM
software (e.g. [15]), we analyze the generated NC code
to make specific predictions of electricity requirements
in manufacturing for the selected process conditions.
This method provides case-specific feedback about the FIGURE 1. User interface for providing toolpath parameters after
environmental impact of the part’s manufacturing, since the uploading an NC code
same part could be manufactured by a number of alternative
methods/toolpaths, each possibly requiring different time
and energy.
3. Block-by-block: We parse the provided NC code block-
by-block to determine the time and energy requirements Energy analyzer process flow
for each block separately using the energy model. The The energy analyzer software is an embodiment of our
cumulative sum of all blocks in the CNC code gives the total energy consumption model described in the previous section.
time and energy requirement of the CNC code. This method The software accepts an ASCII text file NC code as its input. The
also allows us to visualize detailed analysis of the code for provided NC code is first parsed to detect all the tools used in the
identifying the operations and their environmental impacts complete operation using tool change command instances. Then,
individually. for each detected tool, the user is asked to specify a tool diameter
4. Web-based implementation: We have implemented the and approximate overlap percentage that is later used in energy

4 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

FIGURE 2. The output screen showing the expected time, energy consumption, GHG emissions, and block-by-block plots

computations. The user may also specify other parameters that two plots indicating expected block-by-block time and energy
were used when generating the NC code in the CAM software. consumption (Fig 2).
On providing the tool diameters and CAM software
parameters, the main execution loop of the analyzer begins
parsing the NC code for the second time, processing one block RESULTS
at a time, until the end of file is reached or a terminate statement Validation of energy analyzer predictions
code (M30) is read. During execution, the three power states of We performed additional machining experiments to validate
Pconst , Pmove , and Pcut in the proposed energy model (Equation 11) the energy analyzer predictions. The objective of these
are continually updated by the software. For each block, Eblock experiments was both to compare the total energy use as
is estimated from Pblock and tblock using Equation 3. The total predicted by our software against the observed value, as well
energy and time estimates are updated after each block by adding as to compare values of various power states predicted by the
the Eblock and tblock values. In the next section, we illustrate the software against those observed instantaneously on an energy
user interface of the software. meter during actual machining.
The data presented here reflects values averaged over three
(or more) trials for each experiment, since we want to avoid
Software implementation and interface effects of fluctuations in the electricity supply. These validation
We have implemented the energy analyzer in a client-server trials were performed for a Haas VF-0 3-axis CNC machine
architecture so that it can be accessed in a browser using any in the Mechanical Engineering student machine shop at UC
capable device. The server-side scripting is done in PHP. The Berkeley. In all trials, we used Aluminum 6061 workpieces and
software internally uses the Google Charts API [16] to plot time 2-fluted uncoated HSS end-mill tools of 0.375 inch diameter.
and energy estimate graphs in a browser during output display. We used a Yokogawa CW240 energy meter to record electricity
Analysis of a typical NC code starts with uploading the NC code consumption data such as the voltage, current, active and reactive
to the energy analyzer web-portal. Figure 1 shows the input power, and power factor for each phase of the 3-phase electricity
screen that appears after an NC code is uploaded. On this page, supply. Also, to differentiate validation experiments from
tool details, the starting z-level, and the typical programmed customization experiments, we measured various power states
overlap in the uploaded NC code are obtained from the user. The of the machine at different values of input parameters than those
user must provide tool diameters for each tool used in the NC used in the trials to compute the customization constants. Results
code. from the two types of validation trials are presented in the next
two sub-sections.
With these inputs and the NC code, the energy analyzer then
computes the expected machining time (in seconds), required
energy (in W-sec), and green-house gas (GHG) emissions (in Validation of various power states
Kg) based on United States Environmental Protection Agency In the first set of validation trials, we validated the
(USEPA) data that relates the energy consumption to GHG intermediate power states of the CNC machine as predicted by
emissions for the 50 states [17]. The output screen includes the energy analyzer. In all these trials, we used the manual data

5 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

input (MDI) mode of the CNC machine in order to measure a trends in the validation checks were consistent with those found
particular power state. By using MDI mode, it was possible in the machining trials.
to isolate power consumption easily because unlike a program
run, the machine stops after each MDI command. In Table
Case Studies
1, the power states predicted by the energy analyzer and those
We now present three case studies to highlight the
measured by the Yokogawa meter are compared. Deviation from
importance of being able to predict the electricity usage for
the observed power state values was within 1-6%.
toolpaths (G/M codes). In all test cases presented, two alternative
toolpaths for the same component are analyzed.
Validation of complete NC codes (toolpaths)
In the subsequent validation trials, we measured the energy Case study 1: Facing operation alternatives
consumption of various complete NC codes (toolpaths) and Objective: In this case study, we use the energy analyzer
compared the measured values with those predicted by the to compare the energy requirements of two nearly-identical
energy analyzer. These trials were performed in the (usual) alternative toolpaths for facing a square block that only differ
program mode with a Yokogawa CW240 meter connected to the in the direction of cut.
machine to measure the electricity consumption of the complete
NC code. All machining parameters other than the toolpaths
were held constant in each validation run as well as the test cases
presented in the following section (e.g. depth of cut = 0.05 in).
Toolpaths used in these validation trials are shown schematically
in Figure 3. In Table 2, results of the validation trials for
6 complete NC codes are presented along with the calculated
percentage deviations. Similar to customization constants, each
measured power state is an average power state value of the three
independent observations in the square brackets.

FIGURE 4. Face-milling a square face with cutting primarily in (a)

X-axis direction (b) Y-axis direction

FIGURE 3. (a-f) Schematics of the six toolpaths used in validation Procedure: Commercial CAM software programs (e.g.
checks. MasterCAM) provide many alternative strategies for facing; we
consider the zig-zag machining strategy for this case study, due
to its popularity as the strategy of choice in rough machining
We found that our energy analyzer successfully predicted operations such as facing [18]. We deliberately generate NC
the time and the energy usage for every toolpath within a range codes from two toolpaths such that other than the primary
of -6% to +5% and that the trends predicted by the energy direction of cut, all other geometric and mechanical parameters
analyzer (e.g. the energy usage of X-directional facing smaller of the two toolpaths being compared are identical. Toolpath A,
as compared to the Y-directional facing, the energy usage of seen in Fig. 4a, has the majority of cutting in the X-axis parallel
spiraling in smaller than that of that of spiraling out) also direction while toolpath B, seen in Fig. 4b, has the majority of
matched with the observed values in machining trials. The cutting in the Y-axis parallel direction. Since this case study is for
deviation of the estimates from the observed values was found a square block, both toolpath alternatives have nearly identical
on both sides of the observed values. machining time for the same cutter diameter and overlap. Results
We believe that the results of the validation checks are from running the energy analyzer for the two toolpaths (using
promising and that the energy analyzer can be used with a NC codes generated by MasterCAM) are presented side-by-side
reasonable faith to estimate time and energy consumption of NC in Table 3.
codes without having to actually execute the codes on a CNC Observations and inferences: We notice that, although the
machine. We further propose that the energy analyzer can also expected machining time for the two toolpaths is nearly identical
be used to rank different toolpaths against each other since the (324 vs. 318 seconds), toolpath A, where most of the cutting

6 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

TABLE 1. Validation of power states
Power State Estimate (W) Avg. Measured (W) [T1, T2, T3] Deviation (%)

Constant Power (tare, coolant pump) (Pconst ) 325 323 [323, 320, 325] -0.6
Spindle Power (Pspindle at 5500/6700 rpm) 1257/1526 1270 [1290, 1260, 1260]/1540 [1530, 1540, 1550] -1/-1
X-direction Pmove , 25%/60% Rapid 296/361 310 [310, 310, 310]/380 [390, 370, 380] -4/-5
Y-direction Pmove , 25%/60% Rapid 365/388 360 [360, 360, 360]/400 [390, 400, 410] 3/-3
X-direction Pcut , Feed 30/50 IPM 314/347 300 [300, 290, 310]/370 [370, 370, 370] 5/-6
Y-direction Pcut , Feed 30/50 IPM 402/429 430 [420, 440, 430]/440 [460, 430, 430] -6/-2

TABLE 2. Validation of NC codes (toolpaths)

Time (s) Energy (W-sec) Deviation (%)
Estimated Measured Estimated Measured Time Energy

a. Component 1 (Facing, X directional) 286 290 109,876 106,475 -1 3

b. Component 1 (Spiraling in) 262 265 109,223 114,542 -1 -4
c. Component 1 (Spiraling out) 267 280 123,871 130,652 -4 -6
d. Component 1 (Facing, Y directional) 291 287 147,190 142,601 1 3
e. One contour of a face 17.2 17.3 28,407 27,913 0 1
f. Sides of a face contour 15.6 17 28,012 27,633 -2 1

happens along the X-axis, is estimated to require about 20% less is remarkable and entirely consistent with the measured
energy than toolpath B, where most of the cutting happens along differences in the validation experiments shown in Table 1.
the Y axis direction. Based on the observations in the previous Moreover, for a much larger industrial CNC machine such as a
sub-section during power-states validation, we anticipate that due Heynumill 3200 PF [19] that has a gantry-type configuration, the
to the differences in axis construction, motor capacities, and the two principal axes can be expected to have significantly different
bearing sizes, the amount of energy to move or cut in the X- energy consumption properties because one of them carries the
axis direction is more than that in the Y-axis direction. The Haas additional mass of the spindle and the tool changer magazine
VF-0 machine has the X-axis mounted on top of the Y-axis as (the gantry-styled axis) while the other (the table) carries the
seen in Figure 5. As a result, when machining in the Y-axis additional mass of the work-piece itself (7,500 Kg against 57,000
direction, the whole mass of the X-axis motors, bearings, guides, Kg). Further, the axis motors used to move the axes have power
etc. must move, which may be the reason for the difference in ratings of 9.7 KW against 34.2 KW. Thus, in facing a rectangular
energy requirements of the axes. Both toolpaths require nearly
identical machining time (the small difference is partly due to the
difference in starting/ending positions and fixturing) due to the
square geometry of the face and the identical zig-zag toolpath
strategy in both cases.

TABLE 3. Comparison of estimates for facing operation alternatives

Toolpath Energy (W-s) Time (s)

X-directional facing 106,241 324

Y-directional facing 130,497 318

Significance: The predicted difference (>20%) in energy FIGURE 5. Axes mounting in Haas VF-0 indicating X-axis being
consumption for identical movements in different axis directions mounted atop Y-axis

7 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

face on this machine, machining along the first direction may TABLE 4. Comparison of estimates for spiral and contour parallel
be significantly more energy-efficient when compared to facing toolpaths
along the other direction. Toolpath Energy (W-s) Time (s)
Case study 2: Effect of “corners” in toolpath
Objective: In this case study, we compare two toolpath Contour-parallel 88,144 234
strategies that differ markedly in the number of corners. True spiral 78,896 219
Procedure: We compare the expected energy consumption
of two toolpath alternatives for facing a square block using two
entirely different strategies that are popularly used in industry.
For facing a square block, in MasterCAM, we select a contour- Procedure: We compare the expected energy consumption
parallel toolpath and a spiral (inwards) toolpath with the same of two toolpaths alternatives for facing the input geometry
percentage overlap. The toolpath based on contour-parallel with islands using the same two strategies used in the
strategy has significantly more 90-degree turns or corners (Figure previous example. We again generate the required toolpaths in
6a), while the one based on the spiral strategy has virtually no MasterCAM, one with a contour-parallel toolpath strategy and
sharp turns (Figure 6b). the other with a spiral (inwards) toolpath strategy with the same
Observations and inferences: Results from running the percentage overlap. The toolpath based on the contour-parallel
energy analyzer for the two toolpaths (using NC codes generated strategy has significantly more 90-degree turns or corners (Figure
by MasterCAM) are presented side-by-side in Table 4. We find 7a), while the one based on the spiral strategy has virtually no
that the spiral strategy is not only more efficient in terms of sharp turns but has many lifts resulting from the islands (Figure
machining time (approximately 7%) but may also save energy 7b).
(10%) during machining. This behavior may be attributed to Observations and inferences: Results from running the
frequent and sudden accelerations or deceleration of the axes energy analyzer for the two toolpaths (using NC codes generated
motors (at corners) that results in higher instantaneous power by MasterCAM) are presented side-by-side in Table 5.
consumption for the contour-parallel path. Furthermore, [20] Significance: We now find that the spiral strategy is
report that a toolpath with numerous corners, as found in approximately 10% less efficient in terms of machining time
the contour-parallel toolpath in this case-study, may result in and energy consumption as compared to the contour parallel
undesirable tool marks left on the machined surface due to strategy for this input geometry. This is in stark contrast to the
sudden acceleration of the axes. earlier example where the spiral strategy was found to be more
Thus, we can tentatively claim that corners in a toolpath are energy efficient as compared to the contour-parallel toolpath
not favorable for energy efficiency since they represent lost time strategy. Thus, we find that although corners are worse for
and energy in starting and stopping an axis motor. energy consumption of a toolpath (from the earlier example),
the input geometry also plays an important role in deciding a
strategy for machining a component and that no one strategy can
be guaranteed to be energy efficient for all input geometries.
TABLE 5. Comparison of estimates for spiral and contour parallel
toolpaths for a face with islands

Toolpath Energy (W-s) Time (s)

Contour-parallel 109,874 311

True spiral 121,710 349

FIGURE 6. (a) Contour-parallel (b) Spiral toolpath for a square facing The three cases explained above illustrate possible use-cases
operation for the energy analyzer as a supporting software to MasterCAM
Case study 3: Effect of input geometry or any other toolpath planning software to make manufacturing
Objective: In this example, we highlight the importance engineers aware of the effects of process planning choices on
of input geometry when deciding a strategy for machining. We energy costs. Although the analyzer’s predictions were validated
again compare the same two toolpath strategies (one with and the for both precision and trends, we identify here some of its
other without sharp corners) to generate toolpaths for a different limitations that arise from the data acquisition process or the
input face geometry: a rectangular face with two islands. equipment available for testing in a university machine shop.

8 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

power supply, effects due to the presence of other electric
equipment in the vicinity, and the relatively low frequency of
measurement of the energy meter – the Yokogawa CW240
(10Hz). Ideally, the energy meter must be integrated with
the CNC controller during measurements. A WattNode
energy meter [22] may be suitable for this purpose.

With the ultimate aim of developing knowledge and usable
tools that can aid in making energy-conscious process planning
decisions, we described in this paper the development of an
analytical model for CNC machining that correlates several
process parameters to the energy consumed in the process.
Our model is novel because in addition to the known factors
(MRR and process time), it also considers geometric aspects of
toolpaths and machine construction information when making its
FIGURE 7. (a) Contour-parallel (b) Spiral toolpath for a facing predictions. The model was validated using data from machining
operation for a face with islands experiments and we found that the model, while dependent
on the accuracy of the apparatus and conditions, is reasonably
accurate (always within -6 to +5%). We also described an
Limitations application of our model in the form of an energy analyzer
1. The time computations in the energy analyzer assume software tool that could be readily used on the shop-floor by
a linear acceleration and deceleration profile of the axis process planners to compare and contrast energy consumption
motors and an almost universal value of 1m/s2 for the requirements of the various toolpath options available from a
acceleration of an axis. These computations can be further CAM software. We believe that with this model and software,
improved with more accurate acceleration and deceleration informed decisions can be quickly made on the shop-floor that
profiles for each machine. will ultimately help in saving energy in the CNC machining
2. The energy estimates in the energy analyzer are based on processes.
an inherent assumption that up-milling and down-milling
require the same amount of energy if all other conditions ACKNOWLEDGEMENT
are identical. This, however, is only an approximation. We are grateful to Yusuke Yasui who implemented the
Theoretically, down-milling requires less energy but we original version of the energy analyzer software and Anthony
could not discern clear trends in our experiments and thus Bailey Jr. for help in conducting machining experiments. We
were unable to confirm the difference in power consumption. also thank Gordon Long and Dennis Lee in the M.E. student
3. The energy estimates in the energy analyzer rely on a machine shop at UC Berkeley for their help in machining
constant depth of cut in a machining pass. Therefore this experiments. Finally, we wish to thank Dr. Wolfgang Bloehs
energy analyzer is more suitable for analyzing rough-cutting and his staff at AUDI AG for their feedback and providing
operations where more material is removed at a higher data on CNC machines used in the Audi plant at Ingolstadt,
MRR. Our implementation also assumes a flat end-mill. Germany. We particularly thank one of the anonymous reviewers
Minor modifications to accurately compute the cut volume for detailed suggestions.
may be necessary in the software implementation for other
end-mill types. REFERENCES
4. Avram et al. in [21] recommend that in order to [1] Duflou, J., Kellens, K., and Dewulf, W., 2011. “Unit
identify stable cutting conditions when measuring energy process impact assessment for discrete part manufacturing:
consumption by a CNC machine, a modal analysis must be A state of the art”. CIRP Journal of Manufacturing Science
performed. Based on the frequency response function of the and Technology, 4(2), pp. 129–135.
machine tool, stability lobe diagrams should be developed [2] Gutowski, T., Murphy, C., Allen, D., Bauer, D., Bras, B.,
to ensure that during the acquisition, no oscillation behavior Piwonka, T., Sheng, P., Sutherland, J., Thurston, D., and
or any other undesired effects occur that may affect the Wolff, E., 2005. “Environmentally benign manufacturing:
measured electric signals of voltage and current. observations from japan, europe and the united states”.
5. The data acquisition process suffers from fluctuations in Journal of Cleaner Production, 13(1), pp. 1–17.

9 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use:

[3] Franklin, J., Gibbs, K., Stringer, C., Coster, D., Ross, [15] Dassault Systems, 2013. Solidworks 2013. http://
D., Kim, D., Su, C., Maloney, M., and Gillow, E., 2012.
Machining process and tools, Aug. 2. US Patent App. [16] Google Inc., 2010-2014. Google ChartsAPI. https://
[4] Gutowski, T., Dahmus, J., and Thiriez, A., 2006. [17] U. S. Energy Information Admin., 2008. Annual Energy
“Electrical energy requirements for manufacturing Review, Rep. DOE/EIA-0384(2008). [Online; accessed
processes”. In 13th CIRP international conference on life 19-July-2008].
cycle engineering, Leuven, Belgium, Vol. 5, pp. 560–564. [18] Lasemi, A., Xue, D., and Gu, P., 2010. “Recent
[5] Kara, S., and Li, W., 2011. “Unit process energy development in CNC machining of freeform surfaces: A
consumption models for material removal processes”. state-of-the-art review”. Computer-Aided Design, 42(7),
CIRP Annals-Manufacturing Technology, 60(1), pp. 641 – 654.
pp. 37–40. [19] Friedrich Model and Mold GmbH, 2015. Heynumill
[6] Diaz, N., Redelsheimer, E., and Dornfeld, D., 3200 PF.
2011. “Energy consumption characterization and heynumill-3200-pf.
reduction strategies for milling machine tool use”. In [20] Chen, H.-P., Kuo, H.-H., and Tsay, D.-M., 2009.
Glocalized Solutions for Sustainability in Manufacturing, “Removing tool marks of blade surfaces by smoothing
J. Hesselbach and C. Herrmann, eds. Springer Berlin five-axis point milling cutter paths”. Journal of Materials
Heidelberg, pp. 263–267. Processing Technology, 209(17), pp. 5810 – 5817.
[7] Newman, S., Nassehi, A., Imani-Asrai, R., and Dhokia, [21] Avram, O. I., and Xirouchakis, P., 2011. “Evaluating the
V., 2012. “Energy efficient process planning for CNC use phase energy requirements of a machine tool system”.
machining”. CIRP Journal of Manufacturing Science and Journal of Cleaner Production, 19(6–7), pp. 699 – 711.
Technology, 5(2), pp. 127 – 136. [22] Continental Control Systems, 2009-2014. Wattnode
[8] Mori, M., Fujishima, M., Inamasu, Y., and Oda, Y., 2011. pulse meter.
“A study on energy efficiency improvement for machine WattNode_Pulse.
tools”. CIRP Annals - Manufacturing Technology, 60(1),
pp. 145 – 148.
[9] Balogun, V. A., and Mativenga, P. T., 2013. “Modelling
of direct energy requirements in mechanical machining
processes”. Journal of Cleaner Production, 41(0), pp. 179
– 186.
[10] Altintas, Y., 1992. “Prediction of cutting forces and tool
breakage in milling from feed drive current measurements”.
Journal of Engineering for Industry (Transactions of the
ASME), 114(4), pp. 386–392.
[11] Balogun, V., Aramcharoen, A., Mativenga, P., and
Kah Chuan, S., 2013. “Impact of machine tools
on the direct energy and associated carbon emissions
for a standardized nc toolpath”. In Re-engineering
Manufacturing for Sustainability, A. Y. C. Nee, B. Song,
and S.-K. Ong, eds. Springer Singapore, pp. 197–202.
[12] Pavanaskar, S. S., 2014. “Improving energy efficiency in
CNC machining”. PhD thesis, University of California,
[13] Smith, S., and Tlusty, J., 1991. “An overview of
modeling and simulation of the milling process”. Journal
of Engineering for Industry (Transactions of the ASME),
113(2), pp. 169–175.
[14] Layek, R., Nounou, H., Nounou, M., Datta, A., and
Bhattacharyya, S. P., 2012. “A measurement based
approach for linear circuit modeling and design”. In
Decision and Control (CDC), 2012 IEEE 51st Annual
Conference on, IEEE, pp. 3821–3826.

10 Copyright © 2015 by ASME

Downloaded From: on 03/16/2018 Terms of Use: