2212-8271 © 2015 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 organizing committee of CIRPe 2015 - Understanding the life cycle implications of manufacturing doi: 10.1016/j.procir.2015.08.063
Procedia CIRP 37 ( 2015 ) 18 – 23
ScienceDirect
CIRPe 2015 - Understanding the life cycle implications of manufacturing
Increasing energy flexibility of manufacturing systems through flexible
compressed air generation
Jan Beier
a,*, Sebastian Thiede
a, Christoph Herrmann
aaSustainable Manufacturing and Life Cycle Engineering Research Group, Institute of Machine Tools and Production Technology, Technische Universit¨at Braunschweig, Langer Kamp 19b, 38106 Braunschweig, Germany
∗Corresponding author. Tel.:+49-531-391-7153; fax: +49-531-391-5842. E-mail address: j.beier@iwf.tu-bs.de
Abstract
A continuing increase in energy supply from variable renewable energy (VRE) sources requires new strategies to match energy supply and demand. Demand response (DR) strategies focus on flexibilizing energy demand. In this context, manufacturing systems can be designed and controlled to achieve a better fit between energy demand and volatile supply. To maximize energy flexibility, not only manufacturing processes but also auxiliary systems such as compressed air (CA) supply offer opportunities for DR actions. Nonetheless, dynamic behavior and dependencies between manufacturing processes, auxiliary services and resulting overall energy demand requires an integrated approach. This paper presents a method to control production systems and CA supply to increase energy flexibility while maintaining manufacturing system throughput and considering dynamic system dependencies. A production system with several processes, buffers and CA supply system is modeled and simulated in a mixed continuous-time and discrete-event environment. Energy control strategies are implemented and their effectiveness is evaluated. A case study is used to demonstrate that VRE integration can be improved through process and CA supply control without compromising throughput. A focus is set on CA supply and its influence on energy flexibility: the effect of increased CA system volume and additional compressor capacity is investigated.
c
2015 The Authors. Published by Elsevier B.V.
Peer-review statement : Selection and peer-review under responsibility of the International Scientific Committee of the 4th CIRP Global Web Conference in the person of the Conference Chair Dr. John Ahmet Erkoyuncu.
Keywords: Manufacturing system simulation; energy flexibility; demand response
1. Introduction
In 2010, industry was the largest contributor of direct and in-direct annual greenhouse gas emissions of all sectors. Industrial emissions summed up to 28.6% of the global 49.5 GtCO2eq/yr
emissions, of which 18% were direct emissions and 10.6% in-direct emissions through electricity and heat production [1]. As a result, reducing emissions from industry plays a vital role for mitigating climate change. Aside from enacting measures ad-dressing energy intensity and efficiency of industry, low carbon energy and electricity supply such as renewable energy (RE, e.g. biomass, solar/wind energy) are regarded as central strate-gies for reducing overall carbon emissions and increasing en-ergy independence [2]. Some of these RE sources are so called variable renewable energy (VRE) sources (e.g. wind). VRE sources are characterized by temporarily changing availability, non-dispatchability and largely decentralized generation capac-ities (both RE and VRE) [2].
A special challenge emerges when VRE sources start to sup-ply a substantial share of electricity within a power system (e.g.
in Denmark or Germany) or if locally generated VRE ought to be directly consumed (decentralized generation) to avoid trans-portation requirements. As VRE output can change signifi-cantly and electricity demand can be relatively inelastic, solu-tions have to be found to overcome local and temporal demand-supply mismatches. In the context of high VRE penetration, two general strategies are (1) storing electricity for later demand or (2) reshaping demand to better match supply, also known as demand response (DR) [3]. Storing electricity for later use generally reduces efficiency due to conversion inefficiencies or might simply be unavailable (i.e. pumped hydropower requires suitable geological characteristics). Reshaping demand can be a viable strategy as inefficiencies might be lower. Further, a flexible demand enables additional strategies such as pricing arbitrage (shifting electricity demand from high prices to times with low prices), control and reduction of peak demand or par-ticipation in reserve capacity markets. In the context of unstable or unavailable grid supply, energy flexibility can contribute to an energy-autarkic production. However, restrictions need to be considered (i.e. fulfilling production targets, area illumination if
© 2015 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/).
3.1. Electricity supply
Electricity is assumed to be generated on-site from solar and wind resources. On-site generated electricity is always utilized first (if available and demanded), grid electricity is only utilized if on-site capacities are insufficient for supply. Surplus own generation is fed back into the grid.
3.2. Real-time integrated control strategy
The control strategy aims at maximizing demand of on-site generated electricity. The amount of own available electricity is determined and processes are scheduled to yield the best elec-tricity demand fit with own elecelec-tricity supply. In order to re-flect different control behavior of processes, processes are ei-ther classified as binary (e.g. a turning machine, a compres-sor) or continuous processes (e.g. a conveyor belt). A binary process is assumed to process a part within a fixed cycle time during which the process cannot stop. A continuous process has a flexible cycle time with a minimum and maximum cy-cle time, production speed can be freely adjusted within these boundaries. All processes in the considered system (manufac-turing, transportation, compressors) are controlled simultane-ously. First, all processes are excluded which are not adjustable (i.e. due to min/max compressed air pressure or being the bot-tleneck process; details on non-adjustability determination can be found in the next section). In a next step, all possible produc-tion/idle combinations for all binary processes are determined and the combination with the closest fit to available electricity minus current demand through continuous processes is chosen, processes are scheduled accordingly. As scheduling of binary processes is unlikely to match available electricity supply per-fectly, rates of continuous processes are adjusted to achieve a better fit. A ranking method for continuous processes is used to prioritize which processes have the largest adjustment poten-tial. Ranking depends on available time until a process blocks or starves a throughput-critical process.
If the system state (adjustable processes, electricity supply) changes, the control cycle is triggered again to match electricity supply and demand considering the new system state.
3.3. System behavior
The system behavior module determines dynamic interac-tion of system components. The manufacturing system is as-sumed to consist of several production machines which are connected through automated transportation systems. Buffers are available between processes and transportation. Each bi-nary process has a given cycle time Ct, related electricity de-mand Elproduce(e.g. in kW) and CA demand CAproduce(e.g. in
Nm3). During idling, the corresponding parameters are Elidle
and CAidle. A continuous process has the same energy demand
parameters, but cycle time can be reduced from positive infinity (corresponding to a zero production rate) to a minimum cycle time Ctmin(maximum production rate). Energy demand is
as-sumed to scale linear with production speed (idle demand plus rate-dependent demand).
The proposed concept aims at achieving a constant, prede-termined throughput (i.e. to ease comparison of different system configurations). To achieve constant throughput, a virtual cus-tomer withdrawals products at the end of the production process
with a constant rate. Processes exchange data between each other to ensure that enough products are available in the last buffer at all times. Utilizing this data, the bottleneck process needs to be determined and maximum production ensured to avoid throughput losses. A process i, with cumulative produc-tion CPi,tuntil model time t (total model time T ) and maximum production rate Rmax
i is the bottleneck process bn at time t if CPbn,t+ Rmaxbn · (T − t) ≤ CPi,t+ Rmaxi · (T − t) ∀i bn. (1)
Maximum production can only be achieved if the bottleneck process sets its production rate to its maximum rate and has no idle time. Therefore all processes ensure that the bottleneck process is never starved or blocked. This includes indirectly influencing processes (i.e. blocking a process which, as a result, blocks the bottleneck). Processes which are throughput-critical are not controlled via the central electricity control.
The CA system consists of several compressors, which re-quire electricity as input to supply CA. Compressors are con-nected to a CA storage tank, which, in turn, feeds the CA distri-bution system. CA pressure is not allowed to drop below pmin
and cannot exceed pmax. Compressors cannot be switched on
by the central electricity control if pmaxhas been reached until
pressure has dropped again, and not be switched off if pminhas
been reached, until a slight pressure increase has occurred.
3.4. Evaluation
The following key performance indicators (KPIs) are used to evaluate the impact and effectiveness of the proposed control approach and system parameter changes:
• Self sufficiency ratio (percent): cumulative own electric-ity demand as a fraction of total cumulative electricelectric-ity de-mand, indicator for utilization of on-site generated VRE. • CO2per product (kg per final product): Estimated carbon
dioxide emissions from electricity demand, assuming zero emissions from renewable energy and 0.595 kg CO2per
kWh grid electricity (German est. average 2013 [16]), in-dicator for carbon footprint estimation of products. • Residence time (minutes): time for a product between
en-tering and exiting the system, arithmetic mean over all fi-nal products, to evaluate how long a product needs to stay in the system for processing.
• Max. compressor switches per hour (count/hour): Maxi-mum number of compressor on-switches per hour for all compressors, to be compared to maximum allowed num-ber of switches.
• Max. external electricity demand (kW): Maximum 15-minute average external electricity demand, might be cost relevant (depending on external supply contract). • Average inventory (workpieces): Number of workpieces
within the system, arithmetic mean over model time, indi-cator for amount of fixed inventory and subsequently re-lated capital cost.
In order to reflect system start and end states, embodied elec-tricity and emissions in intermediate products and compressed air have been excluded from indicator evaluation.
Turning Traub TNS 60 Ct = 263 sec. Elidle= 2.5 kW Elproduce= 4.0 kW CAidle= 0 Nm3/h CAproduce= 0 Nm3/h binary B B B B B B B C u stomer Grinding Studer S40 Ct = 223 sec. Elidle= 9.3 kW Elproduce= 13.0 kW CAidle= 27 Nm3/h CAproduce= 27 Nm3/h binary Transport Automated system Ctmin= 20 sec. Elidle= 0.0 kW Elproduce= 1.5 kW CAidle= 0 Nm3/h CAproduce= 0 Nm3/h continuous Transport Automated system Ctmin= 20 sec. Elidle= 0.0 kW Elproduce= 1.5 kW CAidle= 0 Nm3/h CAproduce= 0 Nm3/h continuous Transport Automated system Ctmin= 20 sec. Elidle= 0.0 kW Elproduce= 1.5 kW CAidle= 0 Nm3/h CAproduce= 0 Nm3/h continuous Milling DMU 100 Ct = 206 sec. Elidle= 4 kW Elproduce= 5.3 kW CAidle= 15 Nm3/h CAproduce= 35 Nm3/h binary
Fig. 2: Case study production process and corresponding simulation parameters (B: buffer).
3.5. Structure and system optimization
The proposed approach offers several options for testing hy-potheses on improving VRE integration and related indicators. As mentioned before, a focus within this study is set on CA supply. Therefore, parameters such as CA storage volume and number of compressors are varied. System evolution is deter-mined and indicators for different input parameters calculated, compared and conclusions drawn.
The following section describes an example application of the concept utilizing a prototypical simulation model and an example production process.
4. Example application case study
The proposed concept has been implemented in a commer-cially available mixed discrete-event and continuous-time sim-ulation environment. For an overview of existing methods on energy demand simulation of manufacturing systems and a con-cept for integrated energy modeling of process chains and fac-tories can be found in [17]. A flexible/mixed simulation envi-ronment is required as continuous and discrete elements are dy-namically interacting. For example, the CA system has contin-uous elements (e.g. CA flows), which trigger (discrete) events such as switching compressors or the electricity control cycle. This behavior would not be adequately reflected in a pure dis-crete model, while a disdis-crete product flow requires an environ-ment with discrete events. First, the case study’s parameters and system structure are described, followed by an analysis of results and their discussion.
4.1. Parameterization and structure
An example process chain is used to demonstrate the ef-fectiveness and application of the concept. The process chain structure, the related workpiece and base case process param-eters can be found in figure 2. Data has been used from typ-ical manufacturing machines. All buffers’ maximum buffer capacity has been set to 300 pieces, initial buffer stock to 45 pieces. Maximum compressed air demand is 62 Nm3/h,
there-fore four compressors with 30/30/15/15 Nm3/h capacity are
installed in the base case (splitting to allow for flexible con-trol, slight overcapacity for operational security and to account for losses). Compressor electricity demand is assumed to be 7 kW/(Nm3/min) for all compressors when CA is supplied
(3.5/3.5/1.75/1.75 kW production demand for the four com-pressors). Idle waiting time is set to 60 sec., during which CA output is zero, electricity demand 10% of production electric-ity demand and a switch to production does not count towards maximum allowed compressor switches. The compressor is
No control Control No control Control P rocesses Compressors
Prod.: one piece flow Comp.: pminto pmax
Prod.: one piece flow Comp.: control el. demand Prod.: control el. demand Comp.: control el. demand Prod.: control el. demand.
Comp.: pminto pmax
Fig. 3: Description of four control strategies evaluated within the case study.
switched off if no production signal is received within the 60 sec. waiting time.
Assuming a production target of six pieces per hour, average CA demand is 48.873 Nm3/h. Using the equation
V B=V1(A− A 2)
ZΔP (2)
from [18], with V B being the required CA tank storage, V1the
CA supply, A = V2/V1with V2the CA demand, Z the max.
allowed compressor switches andΔP the compressor operating pressure band, and assuming a targeted five switches per hour (Z) and an operating band of one bar (ΔP, from seven to eight bar), the resulting CA storage capacity is approx. 4.5 m3.
There-fore, a rounded-up system volume of 5.0 m3is set for the base
case. Losses are assumed to equal 20% of average demand, which is 9.775 Nm3/h.
Wind and solar electricity generation data has been collected from own measurements over a period of 28 days in Septem-ber 2013 (one minute resolution). Supply profiles are scaled to achieve that average wind plus solar supply over total simula-tion time equals average system electricity demand without any control (one-piece flow), with an equal share between wind and solar supply.
4.2. Results
For each compressor configuration set-up, four different control strategies are evaluated (figure 3). A combination of process and/or compressor control is used. No control of pro-duction processes denotes that a pure one-piece flow princi-ple is enacted (i.e. a machine produces one part if one part is withdrawn from its outgoing buffer). No control of compres-sors describes a strategy where comprescompres-sors are switched on if pminhas been reached, and switched off when current tank
pressure equals pmax. As described before, throughput remains
constant for all strategies and parameters. Two different CA system changes are investigated: tank size volume and number of compressors.
4.2.1. Tank size variation
Figure 4 shows the results for six indicators under di ffer-ent CA system volumes and previously described control
strate-0 2 4 6 8 10 60 61 62 63 64 65 66 67
68 Self sufficiency ratio
CA system volume [cbm] [Percent] X 0 2 4 6 8 10 0.78 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96
0.98 CO2 per product
CA system volume [cbm] [kg] X 0 2 4 6 8 10 3000 3500 4000 4500 5000 5500 6000 6500 Residence time CA system volume [cbm] [Minutes] X 0 2 4 6 8 10 0 5 10 15 20 25 30 35
40 Max. compressor switches per hour
CA system volume [cbm] [#/hour] X 0 2 4 6 8 10 21 22 23 24 25 26 27 28
29 Max. external elec. demand
CA system volume [cbm] [kW] X 0 2 4 6 8 10 300 350 400 450 500 550 600 650 700 750 Average inventory CA system volume [cbm] [Pieces] X
No control Only compressor control Only process control Process and compressor control
Fig. 4: Case study results for the KPIs ratio self sufficiency, CO2per product, residence time, max. compressor switches per hour, max. external elec. demand and
average inventory for different CA system volumes (’X’ denotes the base case with five m3system volume and no control).
gies. As a first result, comparing the impact of proposed control strategies with each other and no control, both control strate-gies achieve to increase utilization of on-site generated VRE. Correspondingly, CO2per product is lower as a result of
in-creased VRE utilization. Further, the compressor control alone yields the lowest impact for both KPIs, the process control has a significant higher impact and the combined control the high-est impact of all strategies. This result is consistent with the share of process and compressor electricity demand: the exam-ple process chain requires more electrical energy than energy from compressed air. Therefore, electrical energy demand from processes has a larger potential for load shifting than compres-sor electricity demand. Another noticeable result is the differ-ence between compressor or process control compared to com-bined control: the impact of comcom-bined control is lower than the impact from process control plus compressor control (both sep-arately enacted and impact added). This indicates a marginally decreasing impact of additionally controlled processes. Look-ing into different compressed air system volumes, larger vol-umes improve utilization of on-site generated VRE. However, marginal improvement is decreasing with larger system vol-umes (flatter curve for larger volvol-umes). A certain CA system volume is required to provide enough flexibility (energy stor-age capability) to support the control approach, but as marginal impact is decreasing and larger system volumes usually result in additional effort (i.e. cost for a CA storage tank), the system volume should only be increased to a certain level. In turn, a small CA system volume can effectively hinder the impact of proposed control. This becomes especially evident when evalu-ating further operational KPIs from figure 4. System residence time and average inventory can be significantly lowered with sufficient CA volume as well as compressor operational con-straints such as max. number of switches. Large compressors (3.5 kW) modeled in this case study should not exceed approx. 20 switches per hour, small compressors (1.75 kW) approx. 25 switches per hour [18], which can only be achieved with
suffi-Self sufficiency ratio CO2 per product Residence time Max. compressor switches per hour Max. external elec. demand Average inventory
−10 0 10 20 30 40 50 60 70 80
Relative change to base case [percent]
No control Only compressor control Only process control Combined control
Fig. 5: Relative change to base case for six KPIs and increased compressor capacity: for self sufficiency, a positive change is favorable (increase), while a negative change (decrease) is favorable for all other KPIs.
cient CA system volumes. System residence time and average inventory are lower with larger CA system volumes as more en-ergy can be stored in compressed air than in intermediate prod-ucts. Maximum external electricity demand is fluctuating and on average increasing for larger CA system volumes. Compres-sors have, on average, longer (critical) production times when CA volume is increased, which results in a higher probability that compressors and processes produce at the same time and thus create a higher peak demand.
4.2.2. Number of compressors
As a second application example, effects of an increase in compressor capacity is investigated. System structure and pa-rameters are held constant, CA system volume is set to 5 m3,
and two additional compressors with 30 and 15 Nm3/h are
in-stalled. Relative change compared to the initial setup for previ-ously described six indicators can be found in figure 5. Nearly all indicators are relatively increased. While this is a positive result for self sufficiency ratio, the results for the remaining five indicators are less favorable compared to the base case. Addi-tionally, self sufficiency increases less than the other indicators. Even though additional compressors can charge the CA system faster, which results in a higher utilization of own VRE, idling
electricity demand is higher and thus total energy demand is increased. The next step would be to investigate the effect of removing one compressor.
In summary, a sufficient tank size to support energy flexi-bility (self sufficiency) while maintaining operating constraints and indicators (e.g. max. compressor switches, system resi-dence time) is required. A CA system volume of approx. five to seven m3should be sufficient to obtain good results. A further
increase would result in additional cost (i.e. for a storage tank) and would need to be economically evaluated. Further, for the specific example case, adding more compressor capacity is not suggested. The initial set-up, together with the combined con-trol strategy, yields the most promising results for increasing energy flexibility.
5. Discussion and conclusion
This paper introduced a method for real-time integrated con-trol of compressed air generation and manufacturing processes to increase utilization of on-site generated VRE. Case study results indicate that proposed control method significantly in-creases the match of electricity demand with volatile supply by shifting electric loads and taking advantage of flexibility en-abled by compressed air storage and intermediate product stor-age. Further operational KPIs such as system residence time and maximum compressor switches are evaluated and results can be used to determine a suitable operating strategy and sys-tem parameter changes, e.g. compressed air syssys-tem volume, to improve impact of proposed strategy.
Compared to other existing storage technologies such as bat-teries, pumped hydro or CAES, the proposed approach differs with regards to required conversion cycles and related ine ffi-ciencies: the approach presented here requires no additional conversion cycle as CA is not converted back into electricity and required anyway by production. However, some ine fficien-cies (e.g. higher pressure band) might be present, which have not been quantified in detail but are likely to be lower than losses due to two conversion cycles (see e.g. [14] for an indica-tion on effects on energy efficiency using different pressure lev-els). In addition, an adequate maintenance strategy and general efficiency of the CA system is a prerequisite to enable proposed approach to avoid increased losses due to e.g. a higher pres-sure band or increased compressor wear-out due to additional switches.
As noted before, efficiency of the proposed approach needs to be carefully compared to other storage options such as batter-ies or CAES. The presented approach assumes that some initial flexibility in production scheduling and CA generation exists and thus efficiency can be expected to be relatively high (no ad-ditional energy conversion, both production and CA scheduling are pure load shifting actions). In order to be applied, energy demand needs to be different depending on process states (i.e. if idle and production demand are similar, scheduling has little impact on total demand). While compressors energy demand is, in general, strongly dependent on production rate, produc-tion machines can have various types of energy demand profiles (i.e. a washing machine can have a nearly load-independent de-mand). Thus, efficiency and applicability depends on overall system structure (scheduling flexibility) and process parameters (individual energy demand profiles) and needs to be
individu-ally determined.
Future research will focus on integration of electrochemical storage options and process cooling into the control method. Further, a monetary evaluation will be set-up, which enables economic comparison of strategies and measures. This will al-low relaxation of constant throughput requirement, which is an-other lead to follow. A comprehensive environmental impact assessment, including additional impact caused by altering pro-duction system parameters (e.g. additional CA storage, battery storage) to quantify overall impact/efficiency and different en-ergy storage/load management options is also developed. References
[1] IPCC. Climate Change 2014: Mitigation of Climate Change. Working Group III Contribution to the IPCC 5th Assessment Report: Introductory Chapter. Tech. Rep.; Intergovernmental Panel on Climate Change; 2014. [2] IPCC. Climate Change 2014: Mitigation of Climate Change. Working
Group III Contribution to the IPCC 5th Assessment Report: Chapter 7: Energy systems. Tech. Rep.; Intergovernmental Panel on Climate Change; 2014.
[3] Pina A, Silva C, Ferr˜ao P. The impact of demand side management strate-gies in the penetration of renewable electricity. Energy 2012;41(1):128– 137.
[4] Lund H, Salgi G. The role of compressed air energy storage (CAES) in fu-ture sustainable energy systems. Energy Convers Manag 2009;50(5):1172– 1179.
[5] Madlener R, Latz J. Economics of centralized and decentralized com-pressed air energy storage for enhanced grid integration of wind power. Appl Energy 2013;101:299–309.
[6] Crotogino F, Mohmeyer KU, Scharf R. Huntorf CAES: More than 20 Years of Successful Operation. In: Solut Min Res Inst Spring Meet. Orlando, FL; 2001, p. 351–357.
[7] Zunft S, Jakiel C, Koller M, Bullough C. Adiabatic Compressed Air Energy Storage for the Grid Integration of Wind Power. In: Sixth Int Work Large-Scale Integr Wind Power Transm Networks Offshore Wind. Delft; 2006, p. 1–6.
[8] Bullough C, Gatzen C, Jakiel C, Koller M, Nowi A, Zunft S. Advanced Adiabatic Compressed Air Energy Storage for the Integration of Wind En-ergy. In: Proc Eur Wind Energy Conf EWEC 2004. London; 2004, p. 22–25.
[9] Karellas S, Tzouganatos N. Comparison of the performance of
compressed-air and hydrogen energy storage systems: Karpathos island case study. Renew Sustain Energy Rev 2014;29:865–882.
[10] Nielsen L, Leithner R. Dynamic Simulation of an Innovative Compressed Air Energy Storage Plant - Detailed Modelling of the Storage Cavern. WSEAS Trans Power Syst 2009;4(8):253–263.
[11] Palensky P, Dietrich D. Demand side management: Demand response, intelligent energy systems, and smart loads. IEEE Trans Ind Informatics 2011;7(3):381–388.
[12] Albadi MH, El-Saadany EF. A summary of demand response in electricity markets. Electr Power Syst Res 2008;78:1989–1996.
[13] Saidur R, Rahim NA, Hasanuzzaman M. A review on compressed-air en-ergy use and enen-ergy savings. Renew Sustain Enen-ergy Rev 2010;14(4):1135– 1153.
[14] Kleiser G, Rauth V. Dynamic Modelling of Compressed Air
En-ergy Storage for Small-Scale Industry Applications. Int J EnEn-ergy Eng 2013;3(3):127–137.
[15] Beier J, Thiede S, Herrmann C. Real-time control simulation of energy flexible manufacturing systems under variable renewable energy supply: concept and case study. Submitted to JMSY 2015; under review.
[16] Icha P. Entwicklung der spezifischen Kohlendioxid-Emissionen des
deutschen Strommix in den Jahren 1990 bis 2013. Tech. Rep.; Umwelt-bundesamt; 2014.
[17] Herrmann C, Thiede S, Kara S, Hesselbach J. Energy oriented simulation of manufacturing systems - Concept and application. CIRP Ann - Manuf Technol 2011;60(1):45–48.
