Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 62 (2017) 317 – 322

10th CIRP Conference on Intelligent Computation in Manufacturing Engineering - CIRP ICME '16

Automatic assessment of machine tool energy efficiency and productivity

Matthias Hacksteiner a*, Fabian Duer a, Iman Ayatollahi a, Friedrich Bleicher a
a
Institute for Production Engineering and Laser Technology (IFT), Vienna University of Technology, Austria

* Corresponding author. Tel.: +43-1-58801-31120; fax: +43-1-58801-31199. E-mail address: hacksteiner@ift.at

Abstract

This work presents an approach to determine relevant energy efficiency and productivity KPIs of machining processes based on a real-time
interpretation of sensor data and machine control data. A comparison of the actual power consumption during machining with an energetic model
of the load-free condition enables the calculation of energetic efficiency and primary processing time. The approach was tested on a CNC turning
and milling center equipped with power meters and compressed air sensors. Sensor data as well as relevant machine control data are read,
processed and recorded via SCADA software in order to automatically calculate certain KPIs.

© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
© 2016 The Authors. Published by Elsevier B.V.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Selection and peer-review under responsibility of the International Scientific Committee of “10th CIRP ICME Conference".
Peer-review under responsibility of the scientific committee of the 10th CIRP Conference on Intelligent Computation in Manufacturing Engineering
Keywords: Machine tool; Monitoring; Sensor; Machine control; Data acquisition; Modelling; Energy efficiency; Productivity; OEE

1. Introduction and motivation improve the environmental performance of manufacturing

As manufacturing industries are facing economic challenges
due to increasing global competition, they continually need to
increase productivity while reducing manufacturing costs. The
industrial sector is a substantial consumer of energy and other
resources and thus causes severe environmental impact [1-3].
Therefore, ambitions to reduce the energy intensity of
manufactured products are suitable to enhance both economic
competitiveness and environmental sustainability.
In recent years, different legislative and normative measures
have been taken in order to reduce industrial energy
consumption. ISO 50001, for instance, provides a systematic
approach to continuously improve energy performance and
specifies requirements for process and equipment design,
measurement and documentation [4]. EN 16231, on the other
hand, suggests a methodology for the evaluation of energy data
in order to determine the energy efficiency of certain units
(such as production systems) enabling energy performance
monitoring and a comparison with other units [5]. Energy
efficiency benchmarking is a suitable way to reveal
optimization potentials concerning energy consumption.
Utilized in great quantities, machine tools constitute
substantial industrial energy consumers [6, 7]. Therefore, a
reduction of machine tool energy demand can significantly
processes and thus the CO2 footprint of consumer products.
A traditional product lifecycle consists of three stages:
manufacturing, use and end of life. For a machine tool itself,
the use phase is the most energy intensive phase causing 60%
to 90% of CO2 emissions during its lifecycle [8].
Recently, a draft standard for the environmental evaluation
of machine tools during their use phase was introduced,
presenting a methodology for a reproducible quantification of
energy supplied to the machine in different operating
conditions [9]. Gontarz et al. presented a modular configuration
approach for machine tools based on multichannel
measurements in order to improve energy efficiency and enable
total cost of ownership (TCO) calculations [10].
Several studies have been carried out in order to model the
energy consumption of machine tools and thus to determine the
environmental impact of goods produced [11-16].
The machining time is a key influence factor for the energy
demand of machine tools, especially such with high base load
(i.e. machines with large peripherals such as hydraulic,
machine cooling, exhaust and cooling lubricant systems).
Various studies have shown that high material removal rates
decrease machine tool overall energy consumption when
keeping the volume of removed material constant due to
decreasing machining time [13, 17, 18].

2212-8271 © 2017 The Authors. 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/4.0/).
Peer-review under responsibility of the scientific committee of the 10th CIRP Conference on Intelligent Computation in Manufacturing Engineering
doi:10.1016/j.procir.2016.06.034
318 Matthias Hacksteiner et al. / Procedia CIRP 62 (2017) 317 – 322
Further optimization potential arises from a proper choice of
the tool path strategy during machining [12, 18, 19].
As a conclusion, it is of high importance to use optimal
machining procedures and parameters in combination with
performant tooling systems in order to minimize cycle times
and thus energy consumption. However, over the course of
machining, the process performance might change due to tool
wear, suboptimal machine settings or operating errors. Hence,
it is expedient to monitor certain performance indicators over
time in order to assess and compare different processes.
A typical energy performance indicator (EnPI) used for
benchmarking within or between units is the specific energy
consumption (such as energy per unit produced) [5]. Emerging
trends in such indicators can not only help to validate changes
in energy efficiency but also act as evidence for issues like
process plan deviations and changes in process stability or
quality.
Progress in the field of sensor and data acquisition
technologies enables real-time acquisition and interpretation of
machine tool data. Vijayaraghavan et al. developed an
automated machine tool energy monitoring system using
MTConnect and applied event stream processing techniques to
automate the analysis of energy consumption [20]. Hu et al.
introduced an on-line approach for energy efficiency
monitoring of machine tools via spindle power measurement
based on power balance calculations [21].
Shin et al. presented a predictive analytics model for
machining processes using neural networks [22]. In their work,
big data infrastructure was fed with STEP-NC plan data and
MTConnect machine monitoring data to derive an analytic
model for the machine tool power consumption depending on
cutting parameters.
Bhinge et al. also introduced a machine monitoring system
architecture based on data acquisition via MTConnect [23]. A
data-driven energy prediction model using Gaussian process
regression was developed using power consumption sensor
data as well as operating data obtained from NC-code and
cutting simulation.
Eberspächer et al. presented a model and signal-based
power consumption monitoring concept and approaches to
reduce the power consumption [24]. In their work, machine
control data read via OPC UA and additional sensor data are
used as input for consumption simulation models to provide the
machine operator with detailed power consumption and
distribution data.
This work presents a different machine monitoring approach
and a methodology to determine relevant energy efficiency and
productivity key performance indicators (KPIs) of machining
processes based on real-time interpretation of sensor data and
machine control data. A comparison of the actual power
consumption during machining with an energetic model of the
load-free condition enables the calculation of the energetic
process efficiency and the primary processing time. The
approach was tested on a CNC turning and milling center
equipped with power meters and compressed air sensors.
Sensor data as well as relevant machine control data are read,
processed and recorded via SCADA software and certain KPIs
are automatically calculated, visualized and stored.
2. Experimental setup
In the framework of the research project “eco2production”,
which focused on the development of methods and tools to
enhance energy efficiency and productivity of producing SMEs
[25], an energy monitoring and control system was
implemented in a pilot factory equipped with machine tools.
The electric power and compressed air consumption of these
machines and certain sub-components as well as peripherals is
recorded and visualized. Furthermore, machine control data
acquisition was implemented for one of the machine tools. The
used SCADA software system, SIMATIC WinCC Open
Architecture, supports different communication protocols and
features a SQL-based database.
Fig. 1 shows a diagram of the experimental setup for sensor
and machine control data acquisition for the CNC turning
center EMCO MAXXTURN 45. The machine tool features a
movable counter spindle as well as a tool turret with driven
tools (for milling and drilling operations) and thus seven
individual drives. The machine has a power rating of 25 kVA
and main spindle and counter spindle drive capacities of 13 and
10 kW, respectively.
Figure 1. Diagram of the experimental setup.
Active power is internally calculated by SENTRON PAC
4200 power monitoring devices from measured electric voltage
and current signals. The according data are transferred to the
SCADA system via MODBUS TCP protocol.
Compressed air volumetric flow and pressure are measured
with FESTO SFAB and SDE1 sensors, respectively. The
sensor data are read by SIMATIC S7-1200 PLC and transferred
to the SCADA system using TCP/IP based S7 messaging. The
same S7 protocol is used to read drive data (such as active
power consumption and speed) from the SINUMERIK 840D sl
machine control via according data block addresses.
For both communication protocols (MODBUS TCP and
S7), data transmission is software-driven with polling cycles of
100 ms. The realized temporal resolution of signals read from
the machine control and PLC is around 100 ms on average. The
power monitoring devices, however, deliver new data in a
mean interval of around 200 ms.
 Matthias Hacksteiner et al. / Procedia CIRP 62 (2017) 317 – 322
Table 1 shows an exemplary overview of data read from the
machine control and Fig. 2 a visualization of the acquired data
in the SCADA software system. The current electric power and
compressed air consumption of the machine tool as well as the
power consumption of its cooling lubrication, hydraulic and
drive systems are displayed and plotted. Furthermore, data read
from the machine control such as the operation mode (JOG,
AUTO etc.), the NC program status (program running,
cancelled etc.) is visualized.
Table 1. Exemplary data read from SINUMERIK 840D sl machine control.
type example data block address
operation mode bool JOG DB11.DBX6.2
NC program status bool program running DB21.DBX35.0
speed float z-axis DB250.DBD128F
active power float main spindle DB250.DBD204F
Figure 2. Visualization of the acquired data in SCADA software.
3. Method
3.1. Determination of operating condition and availability
For the given machine tool, following operating conditions
were defined from an energetic point of view: off, standby,
ready for processing and processing. While the operating
condition processing is determined from the NC program status
program running, the others are distinguished via an
interpretation of measured drive system power consumption on
the basis of an energetic evaluation of the machine tool
according to ISO/DIS 14955-2 [9]. On a side note, an approach
was developed to additionally determine the operating
condition processing from measured energy data only by
detecting dynamic changes in the power and/or compressed air
consumption. This enables the assessment of the operating
condition also for machine tools without machine control data
acquisition.
Finally, the availability is simply determined as the share of
the condition processing in a desired timeframe (e.g. one shift).
This lays the basis for the calculation of the overall equipment
effectiveness (OEE) of the machine which can be determined
from availability, effectiveness and quality rate [26, 27].
319
3.2. Determination of process productivity
Fig. 3 shows the measured power consumption of a machine
tool drive system (axes and spindle drives) for exemplary
machining operations performed on a work piece and also
without work piece (i.e. air cutting). According value-adding
and non-value-adding times are highlighted in different colors.
Figure 3. Power consumption of a machine tool drive system for exemplary
machining operations performed with and without work piece.
To determine the productivity of the process ςP, the primary
processing time tpp is set into relation to the total cycle time tcycle
of the machining process (see equation 1). This KPI is suitable
to compare machining processes concerning their productivity
(i.e. the ratio of value-adding time compared to the total
machining time) as low values might act as an indication that
the process design or execution is suboptimal.
tpp
߫ = (1)
tcycle
The cycle time corresponds to the according duration of the
program running signal in condition true. The primary
processing time is defined as the total duration of tool
engagement including preceding and subsequent axis jogging
without chip removal (cf. Fig. 3). Thus, it is the sum of
durations of movement with machining feed rate during a
machining process. In the presented approach, the primary
processing time is determined via an interpretation of the
resulting velocity vres of all axis motions:
vres = ටvx 2 + vy 2 + vz 2 + vz̵ 2 (2)
The primary processing time is increased if the resulting axis
velocity is larger than zero (e.g. standstill) and considerably
smaller than the minimal rapid traverse speed (and, thus,
corresponds to typical machining feed rates).
The proportion of tool engagement time tcut during
movement with machining feed rate acts as another indicator
for inefficient NC programming or machine operation:
320 Matthias Hacksteiner et al. / Procedia CIRP 62 (2017) 317 – 322

tcut
ςcut = (3)
tpp

The tool engagement time is determined via spindle power

consumption. It is only increased if spindle power signals are
in certain ranges, i.e. non-zero and below high values typical
for spindle acceleration processes.

3.3. Determination of process power and efficiency

The power necessary for the cutting process PP is calculated

from the actual power consumption of the active spindle drive
Ploaded and the power consumption of the according drive at the
same speed in unloaded condition Punloaded (see equation 4).
While the former is directly read from the machine control
during machining, the latter is determined via spindle models
Pk(n) depending on the according spindle speed read from the
control. Note that in equation 4, the indices refer to the spindle
type (1 = main spindle, 2 = counter spindle, 3 = tool head).
3

PP = Ploaded ˗ Punloaded = ෍ሾPcontrol k ˗ Pk (ncontrol k )ሿ (4)

k=1

In order to determine the power consumption of the spindle

drives in unloaded condition (i.e. air cutting) as a function of
speed, spindle tests were carried out. In these tests, the speed
of the according spindle was increased from idle state to top
speed in 100 rpm increments with defined hold times at each
speed. During the tests, which were carried out five times for
each spindle, all other drives were kept in idle state. The power
consumption during the hold times was arithmetically averaged Figure 4. Power consumption of main spindle, counter spindle and driven tool
for sensor and machine control data over all tests. head in unloaded condition (sensor data and machine control data).
Fig. 4 shows the results of this evaluation in which the
sensor data was corrected by the respective base load of the Table 2. Modelled power consumption of main spindle, counter spindle and
driven tool head in unloaded condition via polynomial P(n) = a n2 + b n + c.
drive system (i.e. the power consumption of the drive control).
The results show good agreement, especially for the tool head. a b c R2
It is noteworthy, however, that for the main and counter main n ≤ 1800 rpm - 3.59 10-4 - 0.9963
spindle the machine control data scatters strongly at certain spindle n > 1800 rpm 6.76 10-8 3.97 10-5 0.646 0.9991
speed ranges which can lead to uncertainties in the calculation n ≤ 2400 rpm - 3.01 10-4 - 0.9972
counter
of the process power. For this reason, the process power is spindle n > 2400 rpm 4.94 10-8 1.71 10-5 0.723 0.9996
calculated from a moving average value of the control data.
tool head -8.34 10-9 1.41 10-4 - 0.9968
Mathematical fitting was performed on the results in order
to obtain spindle models Pk(n) as a basis for the calculation.
Table 2 shows polynomial fit coefficients of measured spindle Fig. 5 shows a graphical representation of the spindle power
power consumption (in kW). models Pk(n) used for the calculation of the process power.
In order to determine the process efficiency ηP, the process
power is set into relation with the measured total electric power
consumption of the machine Pel as well as an equivalent electric
power due to compressed air consumption Ppn and thermal flow
Pth (e.g. due to external machine cooling, cooling lubricant
processing or exhaustion):
PP PP
ηP = = (5)
Ptot Pel + Ppn ൅ Pth

Figure 5. Modelled spindle power consumption.

 Matthias Hacksteiner et al. / Procedia CIRP 62 (2017) 317 – 322 321

The proportional compressor power is determined from the Therefore, the process energy calculated from machine
compressed air volumetric flow Vሶ and pressure p of the control data via the presented approach could be validated. The
machine tool assuming adiabatic compression (see equation 6). results show excellent agreement with less than 1% deviation.
The efficiency of the compressed air system ηpn was
determined from monitoring data and modelled as a function of
volumetric air flow Vሶ . The equation was simplified by
assuming ambient conditions (Ta and pa) and physical
properties of compressor intake air. Note that with simplified
equation 6, the compressor power is obtained in kW when the
volumetric flow is inserted in Nl/min and the pressure in bar.
κ˗1
Vሶ ρ cp Ta p κ 33 Vሶ (p0.286 ˗ 1)
Ppn = ቎ቆ ቇ ˗ 1቏ ≈ (6)
ηpn pa 1650 + Vሶ

As the considered machine tool is not connected with

external cooling or exhaustion systems, the thermal share in the
total power calculation is omitted. However, it could be
modelled according to equation 7 as a function of flow Vሶ , inlet
temperature Ti and return temperature Tr of the according fluid:

Vሶ ρ cp (Ti ˗ Tr )
Pth = (7)
ηth

To calculate the mean process power and mean process

efficiency for a given machining process, the according signals
are simply averaged over the cycle time. The total energy and
process energy for a given work piece manufactured are
determined via mathematical integration of the power signals:
tcycle
Etot = න Ptot dt = ෍ Ptot i ሺti ˗ ti˗1 ሻ (8)
0 i

tpp
EP = න PP dt = ෍ PP j ൫tj ˗ tj˗1 ൯ (9)
0 j

Fig. 6 shows a schematic flowchart of the calculation

procedure which is executed at each update of the acquired
machine control data (i.e. every 100 ms on average). All
relevant calculated variables are then stored in the database in
order to enable the detection of emerging trends, the
comparison of different processes or the validation of
optimization measures. Figure 6. Flowchart of the calculation program.
Fig. 7 shows a visualization of the calculation results for an
exemplary workpiece machining process on the main spindle.
The calculated process power is visualized in a graph and
compared to the measured machine total and drive system
power consumption. Furthermore, the current operating
condition and the resulting axis speed as well as spindle speeds
are plotted in separate graphs in order to facilitate an
interpretation of the power signals. Finally, the calculated
productivity and energy efficiency KPIs are displayed and
updated as soon as a new process starts (i.e. the operating
condition changes to processing again).
The displayed machining process was repeated without
workpiece (i.e. air cutting) which enabled a calculation of the
process energy from measured drive system power signals (cf.
Fig. 3 and equation 4).
Figure 7. Visualization of the calculation results in SCADA software.
322 Matthias Hacksteiner et al. / Procedia CIRP 62 (2017) 317 – 322
4. Conclusions and outlook
This work presents a monitoring approach for the automatic
and real-time determination of energy efficiency and
productivity KPIs of machine tools. It was tested on a CNC
turning and milling center equipped with power meters and
compressed air flow and pressure sensors. Sensor data as well
as relevant machine control data are read, processed and
recorded using a conventional SCADA software system.
The approach can be used to raise the awareness of machine
tool operators and NC programmers concerning equipment and
process performance and to demonstrate the consequences of
their actions towards energy efficiency and productivity.
Furthermore, it is suitable to benchmark different machine
tools producing similar products and to evaluate machine tool
availability and effectiveness as a basis for OEE calculation.
Future activities will focus on developing a new monitoring
framework for the pilot factory using MTConnect for platform
independent sensor and machine control data acquisition. In
parallel, following previous work [29, 30], OPC UA servers
will be developed for sensor data acquisition and incorporated
into an existing OPC UA server exposing machine tool control
data. Thus, testing the presented approach on other machine
tools (i.e. other machine controls) will be facilitated.
Furthermore, future work could focus on extending the
presented methodology by the last step of OEE determination,
i.e. quality rate. It has been shown that spindle power
consumption increases with increasing tool wear [31].
Therefore, emerging trends in the process power determined
via the presented approach could act as an indicator for tool
wear and therefore work piece quality. According empirical
cutting power models for tool wear monitoring as proposed in
previous studies [32, 33] could be adapted for tool energy input
and implemented into the presented calculation routine.
Following up previous studies [34, 35], the presented approach
could also be extended by machine learning functionalities in
order to enable indirect quality monitoring by predicting tool
wear or surface roughness.
Acknowledgements
The presented research was mainly funded through the
Collective Research Networking (CORNET) project
“eco2production” [25]. The authors gratefully appreciate the
support of EMCO and Siemens.
Commercial systems mentioned in this work should not be
interpreted as recommendation or as implication that these
products are necessarily the best available for the purpose.
References
[1] Hauschild M, et al., From Life Cycle Assessment to Sustainable
Production: Status and Perspectives. CIRP Annals - Manufacturing
Technology 2005; 54(2): 1-21.
[2] International Energy Agency (IEA), Tracking Industrial Energy Efficiency
and CO2 Emissions; 2007.
[3] EUROSTAT, Statistics Explained - Consumption of energy; 2015.
[4] ISO 50001, Energy Management System; 2011.
[5] EN 16231, Energy efficiency benchmarking methodology; 2012.
[6] Gardner, World Machine-Tool Output and Consumption Survey; 2016.
[7] Zein A, Transition Towards Energy Efficient Machine Tools. Springer-
Verlag Berlin Heidelberg; 2012.
[8] Diaz N, et al., Environmental Analysis of Milling Machine Tool Use in
Various Manufacturing Environments. Proceedings of IEEE International
Symposium on Sustainable Systems and Technology 2010; 1-6.
[9] ISO/DIS 14955-2, Machine tools - Environmental evaluation of machine
tools - Part 2: Methods for measuring energy supplied to machine tools
and machine tool components; 2016.
[10] Gontarz A, et al., Framework of a Machine Tool Configurator for Energy
Efficiency. Procedia CIRP 2015; 26: 706-711.
[11] Mori M, et al., A study on energy efficiency improvement for machine
tools. CIRP Annals - Manufacturing Technology 2011; 60: 145-148.
[12] Kong D, et al., Softwarebased tool path evaluation for environmental
sustainability. Journal of Manufacturing Systems 2011; 30: 241-247.
[13] Diaz N, et al., Energy Consumption Characterization and Reduction
Strategies for Milling Machine Tool Use. Proceedings of 18th CIRP
International Conference on Life Cycle Engineering 2011; 263-267.
[14] He Y, et al., Analysis and estimation of energy consumption for numerical
control machining. Proceedings of the Institution of Mechanical
Engineers - Part B: Journal of Engineering Manufacture 2012; 226(2):
255-266.
[15] Calvanese M, et al., Analysis of energy consumption in CNC machining
centers and determination of optimal cutting conditions. Proceedings of
20th CIRP International Conference on Life Cycle Engineering 2013;
227-232.
[16] Aramcharoen A, Mativenga P, Critical factors in energy demand
modelling for CNC milling and impact of toolpath strategy. Journal of
Cleaner Production 2014; 78: 63-74.
[17] Diaz N, et al., Strategies for minimum energy operation for precision
machining. Proceedings of MTTRF Annual Meeting 2009; 47-50.
[18] Guo Y, et al., An operation-mode based simulation approach to enhance
the energy conservation of machine tools. Journal of Cleaner Production
2015; 101: 348-359.
[19] Rangarajan A, Dornfeld D, Efficient Tool Paths and Part Orientation for
Face Milling. CIRP Annals - Manufacturing Technology 2004; 53: 73-76.
[20] Vijayaraghavan D, Dornfeld D, Automated energy monitoring of machine
tools. CIRP Annals - Manufacturing Technology 2010; 59: 21-24.
[21] Hu D, et al., An on-line approach for energy efficiency monitoring of
machine tools. Journal of Cleaner Production 2012; 27: 133-140.
[22] Shin S, et al., Predictive analytics model for power consumption in
manufacturing. Procedia CIRP 2014; 15: 153-158.
[23] Bhinge R, et al., An Intelligent Machine Monitoring System for Energy
Prediction Using a Gaussian Process Regression. Proceedings of IEEE
International Conference on Big Data 2014; 978-986.
[24] Eberspächer P, et al., A model- and signal-based power consumption
monitoring concept for energetic optimization of machine tools. Procedia
CIRP 2014; 15: 44-49.
[25] CORNET, eco2production (ecological and economical production),
www.eco2production.com (accessed 28.01.17).
[26] Singha R, et al., Overall Equipment Effectiveness (OEE) Calculation -
Automation through Hardware & Software Development. Procedia
Engineering 2013; 51: 579-584.
[27] ISO 22400-2, Automation systems and integration - Key performance
indicators (KPIs) for manufacturing operations management - Part 2:
Definitions and descriptions; 2014.
[28] VDI 3321, Optimation of cutting - basics; 1994.
[29] Ayatollahi I, et al., Prototype OPC UA Server for Remote Control of
Machine Tools. Proceedings of International Conference on Innovative
Technologies IN-TECH 2013; 73-76.
[30] Pauker F, et al., Service Orchestration for Flexible Manufacturing
Systems using Sequential Functional Charts and OPC UA. Proceedings
of International Conference on Innovative Technologies IN-TECH 2015;
138-141.
[31] Cuppini D, et al. Tool wear monitoring based on cutting power
measurement. Wear 1990; 139(2): 303-311.
[32] Shao H, et al., A cutting power model for tool wear monitoring in milling.
International Journal of Machine Tools & Manufacture 2004; 44: 1503-
1509.
[33] Yoon H, et al., Empirical power-consumption model for material removal
in three-axis milling. Journal of Cleaner Production 2014; 78: 54-62.
[34] Sick B, On-line and indirect tool wear monitoring in turning with
artificial neural networs: A review of more than a decade of research.
Mechanical Systems and Signal Processing 2002; 16(4): 487-546.
[35] Benardos P, Vosniakos G, Predicting surface roughness in machining: a
review. International Journal of Machine Tools & Manufacture 2003; 43:
83-84.