Parameterized Modeling of the Energy Demand of Machining Processes as a Basis for Reusable Life Cycle Inventory Datasets

Abstract

Manufacturing processes have a significant contribution to energy consumption and related greenhouse gas (GHG) emissions in a product’s life cycle. Today, information on GHG emissions is increasingly demanded from companies in a life cycle perspective, based on the methodology of Life Cycle Assessment. Manufacturing companies supply producers of final products and are, therefore, requested to provide data on GHG of their manufacturing processes and resulting products. Obtaining such data for real-world manufacturing processes represents a huge effort. This challenge can be overcome with the use of a parameterized model, the Extended Energy Modeling Approach (EEMA), that has been developed for the machining process, which is a widespread industrial manufacturing process. The model calculates the total energy demand from power key values, which report the average power consumption of the constant and variable units of the machinery equipment, the consumer groups, as well as the different operating states of the equipment. Therefore, EEMA enables the reuse of a single measurement campaign for follow-up investigations of the specific machine tool, thereby significantly improving the efficiency of data acquisition for the calculation of the total energy demand and life-cycle-based GHG emissions. To use EEMA for the compilation of life cycle inventory datasets, methodological requirements were analyzed to derive a procedure for LCA-compliant datasets for machine tools. The key findings of applying the EEMA for the case study of a turning machine show that the constant consumer groups have a significant influence on the total energy demand. The share of the variable consumer groups in the total energy demand increases with increasing machine utilization but is always below 5%.

1. Introduction

1.1. Background and Motivation

Industrial manufacturing processes account for a large proportion of energy consumption and the related environmental impacts, notably climate change. In 2018, 42% of the global electricity consumption of 23,000 TWh was attributed to the industrial sector [1]. Energy consumption for the machinery sector (IEA countries) was 2260 PJ in the same year [2]. Therefore, assessing and decreasing the energy demand and the related greenhouse gas (GHG) emissions of manufacturing processes are overarching goals for politicians and companies. Today, policies like the EU’s Green Deal and accounting schemes like the GHG Protocol not only address on-site emissions of companies but require the assessment of GHG emissions and reduction measures in a life cycle perspective. Thus, information on manufacturing processes in the supply chain is needed by many companies and stakeholders in the industrial sector.

The methodological basis for a life cycle perspective is Life Cycle Assessment (LCA) as standardized by ISO 14040:2021 and ISO 14044:2021 [3,4], which enables the assessment of the environmental impacts of products, services, and technologies [5]. LCA is based on a process-chain model, where information from all processes in the studied system are aggregated for the study object, e.g., a consumer product like a car. Today, several databases for LCA exist, such as ecoinvent, GaBi, Carbon minds, Agri-footprint, which provide datasets for an LCA study for a multitude of processes like electricity generation or bulk materials. However, when it comes to specific manufacturing processes, data gaps are frequent and the existing datasets are often considered to be outdated and not adequate for the modeling of modern manufacturing processes [6]. Moreover, if information is required in a real-world supply chain, e.g., if a producing company requires data for its specific products from the manufacturing company in the supply chain to fulfill reporting requirements, generic datasets are not adequate, but must relate to the genuine manufacturing of the specific market product. In case of GHG reporting, the most significant contribution often comes from the energy demand (or energy consumption) of a process, which, consequentially, is an essential data requirement for most manufacturing processes.

The collection of data for real production processes poses a huge challenge for manufacturing companies, e.g., from the metal-working sector, where companies are typically suppliers of automotive, engine building, or electrical equipment manufacturers. These suppliers often process a variety of products or workpieces according to their customers’ specifications, which may change frequently. Accordingly, energy consumption has to be assessed for a multitude or products, but measurement campaigns can cause considerable costs (time, money, expertise) and pose the risk of disturbing production processes. Even if the total energy consumption of a process line is known, allocation to the specific product is difficult due to the complexity of many manufacturing processes.

This is specifically true for machining (drilling, cutting, milling), the most frequently applied type of manufacturing process in the metalworking industry. The complexity of the machining process itself with its interactions of different individual aggregates of a machine tool and the uniqueness of the machine tools, which themselves are usually customized custom-made products [7], do not allow general statements about the energy consumption of a process. The total energy demand is governed by multiple contributing factors, including the workpiece, the construction of the machine tool, and the machine utilization. The energy demand, thus, has to be assessed for each individual combination of these factors. Consequently, if primary data of a single measurement campaign is gathered, it should be reused to calculate process variations or to evaluate future outlines of process operations.

In 2012, the CO2PE!-Initiative proposed a systematic approach for generating datasets for energy and resource consumption of manufacturing processes [6]. The approach is based on a structured data collection scheme considering the different operating states of a manufacturing tool. Based on a parametric model, the average energy demand for each operating state can be used to calculate the total energy demand for any sequence of these operating states in a given operating cycle. However, this approach also demands a great effort for data acquisition: it treats the machining processes as a black box and does not account for the specific influencing factors from the workpiece and manufacturing tool. Thus, the generated datasets regarding the energy demand are not transferable. To overcome this problem, the authors proposed the sampling of many individual measurements using their approach and the derivation of generalizable datasets from these empirical data. However, contrary to the authors’ expectation, a broad application of their approach cannot be observed, which can be explained by the high measurement effort in real production plants.

In this publication, the focus was set on the energy demand of machining processes. To tackle the challenges mentioned above, a systematic modular and parameterized approach was developed. This approach allows the use of a single measurement campaign to model the energy demand of process variations on the machine tool, thereby reducing the need for further measurements. The model was named the Extended Energy Modeling Approach (EEMA). To develop EEMA, the literature about energy demand of machining processes and respective models and the methodological requirements for LCA datasets were analyzed. A step-wise procedure was performed to compile EEMA based on Power Key Values (PKV). The results from EEMA were validated with primary data of a real-world machining tool. Furthermore, the future usability of EEMA for the generation of generic datasets for LCA studies and benchmarking purposes was discussed.

1.2. Energy Demand of Machining Processes

1.2.1. Overview

Machining is defined as the separation of material, where layers of material are mechanically separated from a workpiece using the cutting edges of a tool [8]. In this paper, machining processes are understood as all types of cutting operations (e.g., milling), which are carried out on modern machine tools with integrated control technology (e.g., Computerized Numerical Control, CNC). The energy demand of machining processes is—next to the workpiece-specific and technological parameters (e.g., material, feed rate)—dependent on the energy demand of the machine tool. In general, machine tools are complex systems, which are often individually designed according to specific customer requirements. The energy demand of a machine tool is determined by (i) its construction characteristics and (ii) the machine tool’s specific operating states [9]:

(i) The construction characteristics are displayed by various aggregates. The actual cutting unit can be discerned from the auxiliary units, such as computer units and coolant pumps, all of which consume energy for their operation. Following [10], these aggregates are named “consumer groups” in the following. The literature distinguishes between constant and variable consumer groups within the operating state “processing” [11]. Accordingly, constant consumer groups consume a time-independent, constant amount of energy. Their energy demand is not influenced by process or machining parameters (e.g., geometry and cutting depth of the material to be machined). Examples of constant consumer groups include computer units and the lightning [12]. Variable consumer groups have a variable energy demand depending on the specific machining process. Examples are the time-dependent energy demand for material removal and spindle rotation [13]. The selection, composition, and dimensioning of the installed consumer groups have an impact on the amount of energy required [14].

(ii) The operating cycle of a machine tool is composed of various operating states [15]. They range from the warm-up to the off-state and are defined and executed by the machine-specific control system. Different consumer groups are active in each operating state based on the commands assigned by the specific program [16]. If the energy expenditures of the variable consumer groups occur in non-productive operating states, they can be assumed to be constant by the program-related control of the machine tool. Therefore, variable consumer groups actually exist only in the operating state “processing” [15], which has the determining share of the energy demand, especially at high machine utilization (≥80%, which is the aim of an efficiency point of view) [17]. In general, the constant consumer groups dominate the energy demand of the operating state “processing”. They can account for up to 85% [9,11,15]. Ref. [14] estimate the energy demand for the actual cutting operation at only 6.6%, ref. [18] at 7.6%, with respect to the respective case studies investigated.

Two main findings can be derived: First, the energy demand of a machine tool is influenced by constant and variable consumer groups. Consequently, a parameterized approach for the energy demand modeling of machining processes must include parameters to depict the contributions of the respective consumer groups. Second, to fulfill the task of a machining process, the full sequence of all operating states has to be run. Consequently, the assessment of the total energy demand of machining processes has to include the energy demand of the operating state “processing” as well as the energy demand of all other non-productive operating states, including the operating states “warm up” and “off”, proportionally.

This energy demand is named the Total Energy Demand (TED) and it is defined as follows: The TED represents all energy expenditures necessary for executing a machining process, including the energy demand for the constant and variable consumer groups within the operating state “processing” as well as the energy demand of all non-productive operating states of the machine tool within a defined operating cycle proportionally.

1.2.2. Modeling Approaches

Many approaches for modeling the energy demand of machining processes exist, as the review articles show [5,15,19,20,21,22,23,24]. These approaches relate the energy demand to theoretically derived or empirically identified parameters [21]. The latter are often based on empirical constants, i.e., a fixed value measured for a specific machining operation [13]. Since they usually aim at optimizing a specific process, their transferability is limited. Other approaches use theoretical parameters, e.g., the power requirements of individual aggregates, which can also be used to determine the energy demand of other machining processes. The authors often indicate a total energy demand of the specific machining process as the result. However, depending on the subject of investigation, this total energy demand refers to a specific system boundary (e.g., technology or time-related), which is often not clearly defined and does not correspond with the TED defined above. Three categories for the energy demand modeling of machining processes are addressed by existing models (Figure 1):

• Category 1: Consideration of the energy demand of the actual cutting operation as a share of the operating state “processing”. While in some cases, only the energy demand of the actual cutting operation is considered, in other cases, the energy demand of the entire machine tool during the chip removal process is included;
• Category 2: Consideration of the energy demand of the machining process, including the entire energy demand of the machine tool within the operating state “processing”;
• Category 3: Consideration of the energy demand of the machining process, including the entire energy demand of the machine tool within the operating state “processing” plus the proportional energy demand share of all other non-productive operating states within a defined operating cycle.

From these categories, only Category 3 models match the TED as defined above. However, as a drawback, such models [9,25,26,27] do not depict the contribution of the constant and variable consumer groups.

The activity of the consumer groups determines the energy required within the operating states. Once the machine is ready for processing, only additional energy is required for the actual cutting operation. This energy demand is proportional to the production rate [11], resp. the machine utilization. To illustrate the magnitude of this share in comparison to the TED, Figure 2 depicts the energy demand share of the actual cutting operation (a) in the TED. For this purpose, a is also quantified based on the literature-based findings from Section 1.2.1, assuming a machine utilization resp. the share when the machine tool is in the operating state “processing” (z) of 80% within the example.

Figure 3 illustrates the effects on a when z is changed. As can be seen, the energy demand of the actual cutting operation a is not proportional to z.

In conclusion, the actual material removal (e.g., measured as material removal or production rate) is not linearly related to the TED.

1.3. LCA of Machining Processes

1.3.1. Application

Life cycle assessment is a method to quantify environmental impacts “throughout a product’s life cycle from raw material acquisition through production, use, end-of-life treatment, recycling and final disposal” [3,4]. LCA can assist in identifying opportunities to improve the environmental performance of products at various points in their life cycle, informing decision-makers in industry, government or non-government organizations, the selection of relevant indicators of environmental performance, including measurement techniques, and marketing [3,4].

LCAs have been applied to many relevant sectors [28], including industrial manufacturing with machining, as a widespread process. The results of the literature analysis for LCAs applied for machining processes are presented in Appendix B, classifying comparative, non-comparative and method-related LCA studies. Most LCAs are comparative studies, comparing different operating conditions [29,30,31] or cooling lubricant strategies [32,33,34,35,36,37,38,39,40], or comparing machining and additive manufacturing [41,42,43,44]. The non-comparative studies [45] focus on an environmental burden analysis of machining processes, [46] investigate the cross-life cycle phase influences in tool manufacturing, and [47] present a combined LCA hybrid model and real-time monitoring system.

The authors of [36] performed an LCA to investigate lubricating strategies for machining processes including the machining process, flood lubrication oil, flood lubrication emulsion, minimum quantity lubrication oil, solid carbide drill, polycrystalline diamond milling cutter, and identified the machining process as the most significant contributor to environmental impacts.

1.3.2. Methodological Aspects

LCA is conducted in four phases including goal and scope definition, life cycle inventory (LCI), impact assessment, and interpretation. As part of the goal and scope definition, the functional unit (FU) with reference flow and system boundaries have to be defined. The FU displays the “quantified performance of a product system for use as a reference unit” [3], whereas the reference flow represents the “measure of the outputs from processes in a given product system required to fulfil the function expressed by the functional unit” [3]. For LCA studies of machining processes, the FU is often defined as the machined workpiece, which is the ultimate rationale to perform the process of, e.g., cutting. At the same time, this FU is suited for the transfer of the data in the supply chain to be used within an LCA for the final (consumer) product by a producer company. For all reference flows, in general, the result of an LCA in terms of calculated LCI and LCIA result is a linear function of the FU [48].

Another important element of the goal and scope phase is the definition of system boundaries, which specify which parts of a life cycle are included in the assessment. The system boundary depends on the subject of investigation.

Table A4. Literature analysis on the subject of “LCA of machining processes” (sorted by year of publication).

Authors	Comparative LCA	Non-Comparative LCA	Method Development	Scope	Process Under Study	Definition: FU	Definition: Reference Flow 5	Input Flow: Energy Demand 5	Cutting Operation	Operating State—Processing	All Operating States	System Boundary
[31]	x			process comparison: variation of cutting speed, feed rate and axial depth of cut under different cutting fluid supply conditions	turning (In-house cast AXZ911/10SiC metal matrix composites)	technical parameter (Turning process on AXZ911/10SiC MMCs for cutting length of 20 mm with defined cutting parameters)	no	yes	x			gate-to-gate
[44]	x			conventional processes and AM 1	casting, machining, binder jetting, powder bed fusion, bound powder extrusion	product (Double cardan H-yoke)	no	yes		x		cradle-to-grave
[40]	x			FL 2, cryogenic and MQL 3 machining	turning (Ti6Al4V ELI)	cutting time (1 min)	no	yes	x			gate-to-gate (Without considering machine tool, workplace, and components)
[39]	x			two cryogenic cutting conditions: liquid carbon dioxide and liquid nitrogen	drilling (Inconel 718)	technical parameter (One drilled hole having a 5 mm diameter and 5 mm depth)	no	yes	x			gate-to-gate
[30]	x			PCD and conventional cemented carbide WC-Co tools	machining (Titanium alloy) wood working	technical parameter (Machining: surface area generated by one WC-Co tool, resp. 0.01 PCD tool during its lifetime) (Wood working: mass of wood removed by one WC-Co tool, resp. 0.1 PCD tool during its lifetime)	no	yes	x			cradle-to-grave
[38]	x			RHVT-MQCF 4 and MQL	turning (CNC, pure titanium)	technical parameter (Machining of titanium (Grade-2) alloy, 150 mm length and 50 mm diameter (machining surface))	no	yes		x		cradle-to-grave
[29]	x			methodology for evaluation of dry turning process along with optimal machining parameters	turning (Inconel 601 using three turning inserts coated with TiAlN + AlCr2O3 by physical vapour deposition)	cutting time (1 h of CNC longitudinal turning of Inconel 601 workpiece of the following dimensions: 300 mm length and 50 mm diameter)	no	yes		x		gate-to-gate
[36]	x			FL and MQL	drilling (Aluminum, cast iron and steel) milling (Aluminum) milling (Cast iron and steel)	technical parameter (3 drill holes with a twist drill (Diameter 8.5 mm, drilling depth 5xd) and chip volume of 2.411 mm3) (Milling surface of 26.250 mm2 with a cutting depth of 0.2 mm and milling volume of 5.250 mm3) (Milling surface of 2.345 mm2 with a cutting depth of 0.2 mm and milling volume of 469 mm3)	yes 6	yes		x		gate-to-gate
[35]	x			cutting conditions (Dry, mono-jet of cryogenic liquid nitrogen and dualjet of cryogenic liquid nitrogen)	turning (Hardened Ti6Al4V titanium alloy)	not defined	no	yes		x
[46]		x		consideration of cross life cycle phase influences in tool manufacturing	grinding, cutting edge preparation, coating	product (One end mill)	no	yes		x		manufacturing and use phases (Within the theoretical model all operating states are considered)
[34]	x			FL and cryogenic machining	milling (Ti-6Al-4V blank)	product (One Ti-6Al-4V blank)	no	yes			x	cradle-to-grave (Machine tool is considered as black box)
[43]	x			conventional and AddM assisted IC processes	casting (Low-melting alloy) milling (Plaster-like material Aquapour) AM (High Impact Polystyrene) AM (Powder materials)	product (15 mold cores)	no	yes		x		cradle-to-grave
[47]		x		combined LCA hybrid model and real-time monitoring system	grinding	technical parameter (3000 mm3 material removal from a cylindrical workpiece by grinding and with no intermediary wheel dressing)	yes 7	yes			x	gate-to-gate (According to CO2PE!-methodology, including the characterization of machine subunits)
[33]	x			combined techniques based on cryogenic cooling and MQL and other near-to-dry coolant alternatives	turning (AISI 304)	technical parameter (Chip volume obtained by turning a part from Ø59 mm to Ø32 mm using a cutting length of 150 mm)	no	yes	x			gate-to-gate
[42]	x			two AddM machines and a traditional CNC milling machine	AM, milling	product (Two specific parts in acrylonitrile butadiene styrene (ABS) plastic or similar polymer)	no	yes			x	cradle-to-grave (Machine tool is considered as black box)
[37]	x			FL and MQL	machining	product (One bolt: 200 mm length and 42 mm diameter)	no	yes		x		gate-to-gate
[27]			x	case study, demonstrating the application of the screening and the in-depth approach	drilling	technical parameter (Drilling four regularly spaced holes of 19.1 mm diameter through the thickness of the workpiece (50 mm))	yes 8	yes			x	gate-to-gate (CO2PE!-methodology, machine tool is considered as black box)
[41]	x			conventional machining and CLAD-process	milling, drilling, turning, boring, trim die (Ti6Al4V) CLAD-process (Direct additive laser manufacturing)	product (One defined Ti6Al4V mechanical part with a specific technology)	no	yes		x		cradle-to-grave
[49]		x	(x)	inventarisation and analysis of manufacturing unit processes	laser cutting selective laser melting	not defined	yes 9	yes			x	gate-to-gate (CO2PE!-methodology, machine tool is considered as black box)
[32]	x			Near-dry machining and FU machining	gear milling (16MnCr5)	technical parameter (1 kg of alloy steel material)	no	yes	x			cradle-to-grave
[45]		x		environmental burden analyzer for machine tool operations under different cutting fluid conditions	millling	product	no	yes		x		gate-to-gate (Environmental analysis based on emission factors)

¹ Additive Manufacturing (AddM). ² Flood Lubrication (FL), conventional lubrication process. ³ Minimum Quantity Lubrication (MQL). ⁴ Ranque–Hilsch Vortex Tube assisted Minimum Quantity Cutting Fluids (RHVT-MQCF). ⁵ y: yes/considered, n: no/not considered. ⁶ “3 drill holes with FL; 3 drill holes with MQL; Milling with FL; Milling with MQL” [36]. ⁷ “The reference flow used in this study was 3000 mm³ material removal per grinding test (i.e., a complete machining of one workpiece). The UPLCI methodology suggests the use of 1 s of processing time as reference flow […]; however, this may not be the best choice in this case. […], the use of volume of material removed as reference flow enables a better evaluation of how the variation of grinding parameters can impact the environmental performance indicators in an LCA study of manufacturing processes, in comparison to the use of 1 s of processing time” [47]. ⁸ “[…] generally applicable reference flow of 1 s of processing time for a specified load level of a unit manufacturing process for a specified material, based on a working scheme of 2000 h/year […]” [27]. ⁹ “[…] generally applicable reference flow of 1 s of processing time for a specified load level of a unit manufacturing process for a specified material, based on a working scheme of 2000 h/year (250 days with one shift of 8 h) including some specified use modes […]” [49].

(Appendix B) shows examples for system boundaries for the LCA of machining processes.

In the life cycle inventory phase, inventory results are compiled for which life cycle inventory datasets are needed for the processes of the studied system. For the foreground system, i.e., the particular system under study, these LCI datasets are usually compiled by the LCA practitioner. LCI data for background processes, e.g., electricity generation, are often taken from LCI databases. In general, LCI datasets cover material and energy inputs (“inputs from the technosphere”) as well as resources at the input side, and generated emissions, products, and waste at the output side. Input and output data in LCI datasets are compiled from different sources such as direct measurements, process models, bill of materials, or material flow models. An example for an LCI dataset for machining process is given in

Table A2. Modeling differences in the datasets of chromium steel milling.

Ecoinvent Process: Chromium Steel Milling 1, Average/Dressing/Large Parts/Small Parts—RER/RoW
Reference product: chromium steel removed by milling = 1 kg
	average	dressing	large parts	Small parts
Inputs from technosphere
electricity, low voltage	0.67 kWh	11.5 kWh	0.298 kWh	3.19 kWh
compressed air, 700 kPa gauge	1.28 m3	1.28 m3	1.28 m3	1.28 m3
energy and auxiliary inputs, metal working factory	4.41 kg	4.41 kg	4.41 kg	4.41 kg
lubricating oil	0.00382 kg	0.00382 kg	0.00382 kg	0.00382 kg
metal working factory	2.02 × 10−9 unit	2.02 × 10−9 unit	2.02 × 10−9 unit	2.02 × 10−9 unit
metal working machine, unspecified	1.74 × 10−4 kg	1.74 × 10−4 kg	1.74 × 10−4 kg	1.74 × 10−4 kg
chromium steel 18/8, hot rolled	1 kg	1 kg	1 kg	1 kg
Inputs from technosphere, wastes
waste mineral oil	−0.00382 kg	−0.00382 kg	−0.00382 kg	−0.00382 kg
Inputs from environment
water, cooling, unspecified natural origin	0.0148 m3	0.0148 m3	0.0148 m3	0.0148 m3
water, unspecified natural origin	0.00191 m3	0.00191 m3	0.00191 m3	0.00191 m3
Emissions to air
water	0.0063 m3	0.0063 m3	0.0063 m3	0.0063 m3
Emissions to water
water	0.0104 m3	0.0104 m3	0.0104 m3	0.0104 m3

¹ Cf. [57].

. illustrating inputs and outputs for chromium steel milling.

In the life cycle impact assessment phase, inventory results (such as GHG) are translated into environmental impacts. In order to calculate, for example, the global warming potential (GWP) of electricity use for chromium steel milling (example dataset), the amount of electricity is multiplied with the GHG intensity of electricity.

1.3.3. Research Gap

With regard to the application problem introduced in Section 1.1, it is especially relevant to analyze LCA literature on machining processes and developed LCI data sets. A specific focus is set on the question how the FU and reference flow are defined for machining processes and how the input flow “electricity” is modelled.

Available LCI datasets for metalworking machining processes in the ecoinvent (v3.8) and GaBi (2019) database were reviewed and evaluated (Appendix A). Regarding the input flow “electricity” GaBi reflects only the energy demand of the actual cutting operation. In ecoinvent, system boundaries are not defined but it can be assumed that only the energy demand of the actual cutting operation, including the expenses of the machine tool, is contained. Therefore, none of the datasets (

Table A1. LCI datasets of metalworking machining processes in ecoinvent (v3.8, cutoff).

Process Variants						Location			Number of Records
Name	Machine Type	Amount of Material Removed or Process Specification (Input Flow “Electricity”, [kWh]) 1				RER	RoW	GLO
aluminium milling	-	average (0.356)	dressing (6.09)	large parts (0.158)	small parts (1.69)	x	x	-	8
cast iron milling	-	average (0.148)	dressing (2.54)	large parts (0.0659)	small parts (0.706)	x	x	-	8
chromium steel milling	-	average (0.67)	dressing (11.5)	large parts (0.298)	small parts (3.19)	x	x	-	8
steel milling	-	average (0.474)	dressing (8.12)	large parts (0.211)	small parts (2.26)	x	x	-	8
market for aluminium removed by milling	-	average (0.356)	dressing (6.09)	large parts (0.158)	small parts (1.69)	-	-	x	4
market for cast iron removed by milling	-	average (0.148)	dressing (2.54)	large parts (0.0659)	small parts (0.706)	-	-	x	4
market for chromium steel removed by milling	-	average (0.67)	dressing (11.5)	large parts (0.298)	small parts (3.19)	-	-	x	4
market for steel removed by milling	-	average (0.474)	dressing (8.12)	large parts (0.211)	small parts (2.26)	-	-	x	4
aluminium drilling	CNC (0.229)	-	-	-	-	x	-	x	4
	conventional (0.0764)
brass drilling	CNC (0.0625)	-	-	-	-	x	-	x	4
	conventional (0.0208)
Cast iron drilling	CNC (0.167)	-	-	-	-	x	-	x	4
	conventional (0.0556)
chromium steel drilling	CNC (0.75)	-	-	-	-	x	-	x	4
	conventional (0.25)
steel drilling	CNC (0.542)	-	-	-	-	x	-	x	4
	conventional (0.181)
market for aluminium removed by drilling	CNC (0.229)	-	-	-	-	-	-	x	2
	conventional (0.0764)
market for brass removed by drilling	CNC (0.0625)	-	-	-	-	-	-	x	2
	conventional (0.0208)
market for cast iron removed by drilling	CNC (0.167)	-	-	-	-	-	-	x	2
	conventional (0.0556)
market for chromium steel removed by drilling	CNC (0.75)	-	-	-	-	-	-	x	2
	conventional (0.25)
market for steel removed by drilling	CNC (0.542)	-	-	-	-	-	-	x	2
	conventional (0.181)
aluminium turning		average	primarily dressing	primarily roughing	-	x	-	x	12
	CNC	(1.83)	(3.29)	(0.362)
	conventional	(0.347)	(0.561)	(0.134)
brass turning		average	primarily dressing	primarily roughing	-	x	-	x	12
	CNC	(0.992)	(1.79)	(0.196)
	conventional	(0.189)	(0.305)	(0.0727)
cast iron turning		average	primarily dressing	primarily roughing	-	x	-	x	12
	CNC	(1.15)	(2.07)	(0.228)
	conventional	(0.218)	(0.353)	(0.0842)
chromium steel turning		average	primarily dressing	primarily roughing	-	x	-	x	12
	CNC	(2.51)	(4.52)	(0.496)
	conventional	(0.477)	(0.769)	(4.41)
steel turning		average	primarily dressing	primarily roughing	-	x	-	x	12
	CNC	(1.78)	(3.2)	(0.352)
	conventional	(0.338)	(0.545)	(4.41)
market for aluminium removed by turning		average	primarily dressing	primarily roughing	-	-	-	x	6
	CNC	(1.83)	(3.29)	(0.362)
	conventional	(0.347)	(0.561)	(0.134)
market for brass removed by turning		average	primarily dressing	primarily roughing	-	x	-	x	6
	CNC	(0.992)	(1.79)	(0.196)
	conventional	(0.189)	(0.305)	(0.0727)
market for cast iron removed by turning		average	primarily dressing	primarily roughing	-	x	-	x	6
	CNC	(1.15)	(2.07)	(0.228)
	conventional	(0.218)	(0.353)	(0.0842)
market for chromium steel removed by turning		average	primarily dressing	primarily roughing	-	x	-	x	6
	CNC	(2.51)	(4.52)	(0.496)
	conventional	(0.477)	(0.769)	(4.41)
market for steel removed by turning		average	primarily dressing	primarily roughing	-	x	-	x	6
	CNC	(1.78)	(3.2)	(0.352)
	conventional	(0.338)	(0.545)	(4.41)
Total									168

¹ The input flow “electricity” does not change with regard to the location.

and

Table A3. LCI datasets of metalworking machining processes in GaBi (2019).

Process Variants					Number of Records
Name	Amount of Material Removed	Specification	Location	Input Flow “Electricity” [MJ]
aluminium cast part machining	0.02—0.04 kg chips	single route, at plant, specific technology	DE	0.77	4
	0.02—0.3 kg chips			4.42
	complex			2.69
	standard			0.34
Steel cast part machining	not specified	single route, at plant	DE	5.83	1
aluminium machining	19 kg shavings per 1 kg part	single route, at plant, specific technology	DE	34.42	1
cast iron machining	0.05–1 kg chips	single route, at plant, specific technology	DE	0.14	1
steel high-alloyed machining	0.47 kg shavings per 1 kg part	single route, at plant	DE	2.26	1
steel turning	adjustable	single route, at plant	DE	3.30	1
titanium machining	1.86 kg shavings per 1 kg part	single route, at plant	DE	26.89	1
Total					10

) represent the TED. Regarding the reference flow, the ecoinvent datasets define the reference product as the amount of material removed, whereas it is the mass of the machined part in the GaBi datasets. These reference flows cannot provide a linear relation between FU and TED. The CO2PE!-methodology introduced by [6] is the only specific approach for manufacturing and, thus, machining processes. Refs. [47,49] applied the CO2PE!-methodology. Refs. [6,49] distinguished between the FU, which is the product manufactured, and the reference flow. They proposed a time-based “reference flow of 1 s of processing time for a specified load level […] for a specified material” [6] within a working scheme of 2000 h/year. This reference flow integrates the energy demand of the operating state “processing” and of all non-productive operating states proportionally based on primary data (power demand) measured at the machine level. The TED for the provision of the FU is subsequently determined by multiplying the “reference second” with the duration for processing the aimed utility.

As in all studies, generally, the energy demand is considered as input flow; the research framework is, for most studies, on the cutting operation (e.g., [31]) or the operating state “processing” (e.g., [38]) rather than on all operating states (e.g., [34]). Except for [47], all studies that include all operating states within their models consider the machine tool as a black box and, thus, do not distinguish between the constant and variable consumer groups.

As shown above, the CO2PE!-methodology is the only one to account for the assessment of the TED and which provides a linear relation between the FU and the TED. As a drawback, a systematic understanding of the machine tool, the largest contributor to energy demand, cannot be derived since it is treated as a black box. The influencing factors of this energy demand share include the constant and variable consumer groups. However, there is no relation with these specific consumer groups within the CO2PE!-approach [6]. As a consequence, the power demand of the operating state “processing” must be measured separately for each specific machining process (respective product) and no conclusion on the energy demand of other machining processes on the same machine tool can be drawn.

2. Materials and Methods

2.1. Overview on the Methodological Procedure

The overview on EEMA development, validation, and application, as well as the integration into LCI, is illustrated in Figure 4. The procedure is structured in a step-wise approach in which EEMA development is organized in step 1–4. Validation and data acquisition aspects are considered in step 5. The connecting arrow from step 5 to LCI illustrates the link between EEMA and LCI dataset development.

2.2. Development of EEMA

From the state of the art in Section 1.2, the following methodological requirements for a parameterized model were derived:

• The TED must be represented;
• A linear relation between measurable parameters and the TED must be sustained and mathematically provided;
• The model shall allow that a single measurement campaign for a machining process can be transferred to other machining processes on that machine tool.

As the methodological basis for such a parameterized model, the CO2PE!-methodology [6] was chosen, which fulfills the first two requirements. By extending the CO2PE!-method according to [6] with a detailed model for the operating state “processing” considering the constant and variable consumer groups of a machine tool (input model), the Extended Energy Modeling Approach (EEMA) was developed. As part of EEMA, PKV were introduced, which can be easily derived from measured primary data and which allow the fast derivation of the TED. PKV can also be re-used for calculating the TED for other machining processes on the specific machine tool. More precisely, the procedure for the development of EEMA was structured as follows:

• Identification and classification of constant and variable consumer groups using selected information provided by literature;
• Definition of the operating states of a machine tool;
• Selection of a suitable input model for extending the CO2PE!-approach [6] depicting the contributions of the consumer groups within the operating state “processing”;
• Derivation of EEMA based on the PKV;
• The results in terms of a classification system for consumer groups, a definition of operating states, and the developed input and EEMA model are shown in Section 3.1, Section 3.2, Section 3.3 and Section 3.4.

2.3. Validation of EEMA

Step 5 consists of the validation of EEMA and the development of a proposal for data acquisition. For the actual validation of the EEMA, the calculated TED and PKV values per operating state are compared with measured primary data.

To easily implement the model in practice, a procedure for companies was proposed to quickly obtain the data needed for EEMA. The results of validation and the data acquisition proposal are presented in Section 3.5.

2.4. Generation of LCI Datasets

EEMA can furthermore be used to generate LCI datasets. For that purpose, the energy demand based on the PKV, thus representing the TED, is incorporated in the LCI as input flow “electricity”. The reference flow is defined as a reference second of a machining process, derived from EEMA. This reference flow enables a linear relation to the TED. The specific derivation of a reference flow from EEMA is presented in Section 3.6.

System Boundaries

As illustrated in Figure 5 (highlighted in grey), EEMA provides an LCI dataset for machining process that covers the TED of a machining process. This represents a system boundary corresponding to the Category 3 models (see Section 1.2). When this dataset is used in an LCA study to determine, e.g., GHG emissions, it will be linked to an LCI dataset on energy production. Such a study would represent a cradle-to-gate system boundary covering the energy demand and related upstream emissions of the machining process. Other inputs from the technosphere such as lubricating oil, compressed air, etc., are not included in this energy-related system boundary. However, they can be easily added to the LCI dataset of the respective machining process. As demonstrated in [36], the energy demand of the machining process is the largest contribution to GWP as well as most of the other investigated impact categories. The two other relevant contributors, compressed air and lubrication oil, are two independent fields of investigation, which require other methods and data, but whose results can be combined with EEMA.

The presented case studies (Section 3.5 and Section 3.6) focus on the contribution of the energy demand of the machining process exclusively and refer to the partial system boundary covered by the EEMA (grey area).

3. Results

3.1. Classification of Consumer Groups

Consumer groups of machine tools can be distinguished according to their function at the aggregate level within the operating state “processing”. Thus, the classification approaches used by [9,10,11,13,16,17,25,26,50,51] can be compiled as proposed in

Table 1. Consumer groups and classification (constant “c” and variable “v” consumer groups).

Consumer Group	Aggregates, Esp.	Classification
Main switch and control system	Computer, switch cabinet	c
Hydraulics	Pumps	c
Cooling	Cooling pump	c
Suction	Oil mist extraction	c
Lighting	Working room, screen	c
Ventilation	Fans, ventilators	c
Other auxiliary units	Conveyor belt, lubrication	c
Chip conveyor	Motors	c
Cutting fluid supply	Coolant pumps	c
Axis drives	x-, y-, z-Axis	c
Spindle	Motors	v
Tool	Revolver	v
Material removal	Cutting capacity	v

. Consumer groups are classified into the categories constant and variable consumer groups based on the information provided by the respective studies. According to [12], the lubrication and cooling system of some machine tools, e.g., for cooling and lubrication of the spindle, may show periodic changes in the power demand depending on the spindle speed range and the resulting heat generation. Concerning the complexity of the targeted EEMA, the consumer group cooling as well as the consumer group other auxiliary units were assumed to be constant in their average due to the literature-based assignment of the aggregates into the presented categories, the combination of the power demand of the respective aggregates with the power demand of other constant auxiliary units, as well as the uncertain machine tool-related extent of a possible power demand change.

3.2. Definition of Operating States

For the definition of the operating states of a machine tool (

Table 2. Operating states according to their switching states, cf. [10,52].

Operating States	Switching States
	Mains	Machine Control	Peripheral Units	Machine Processing Unit	Machine Motion Unit	Machine Axes
A—Processing	On	On	On	On/P	On/M	On/M
B—Warm up	On	On	On	On/NP	On/M	On/M
C—Ready for processing	On	On	On	On/H	On/H	NM
D—Extended standby	On	On	On	Off	Off	NM
E—Standby	On	On	Off	Off	Off	NM
F—Off	Off	Off	Off	Off	Off	NM
G—Others	On/Off	On/Off	On/Off	On/Off	On/Off	On/Off

M: Moving; NM: Not Moving; H: Hold; P: Processing; NP: Not Processing.

), the classification based on the machine-specific switching states were used as suggested by [52]. Since operating states can occur that cannot be allocated to the operating states A to F, e.g., maintenance operations (cf. [6]), category G, “others”, was introduced.

3.3. Input Model

Category 2 models [13,16,17,50,53] display the parameterized and disaggregated energy demand considering the constant and variable consumer groups. Many of these models use empirically identified constants for a specific case study without including a procedural description of how to determine these constants. Solely, ref. [17] presented a clear description from which measurable parameters for the contribution of the consumer groups can be derived. The authors took the different consumer groups according to the specific load profile of the machine tool into account. By integrating the cutting force, the material, and the cutting depth, the machining characteristics of the workpiece and the associated power demand were considered. Based on [54], they also included the air cutting period because the actual machining time is not only defined by the chip removal but also covers periods in which the tool rotates without having entered the material. Ref. [17] defined three energy shares of the operating state A (E_t, cf. Equation (1)): the energy share of the auxiliary units E_Basic, the energy share of the actual machining process E_Cutting, and the energy share of an intermediate state, the ready state E_Ready. Mathematically, they express this as follows:

(1)

with P_b, P_r, and P_cool [W] as the power demand of the baseload, the power demand within the operating state at the auxiliary unit level and the required power of the coolant pump motor. t_b and t_r [s] represent the periods of the energy states E_Basic and E_Ready, respectively. P_air [W] represents the power demand and t_air [s] the period of air cutting. k is the specific cutting force [kJ/cm³], $\dot{v}$ is the material removal rate [cm³/s], and t_c [s] is the period of the actual cutting operation [17].

3.4. EEMA

The TED of a machine tool TED_M can be described as the integral over the power and can be expressed, within a defined period t, as follows:

$$\begin{gathered} \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{M}}=\left(\sum {\int }_{{\mathrm{t}}_{0,\mathrm{i}}}^{{\mathrm{t}}_{\mathrm{A},\mathrm{i}}}{\mathrm{P}}_{\mathrm{A}}\left(\mathrm{t}\right)\right)+\left(\sum {\int }_{{\mathrm{t}}_{1,\mathrm{j}}}^{{\mathrm{t}}_{\mathrm{B},\mathrm{j}}}{\mathrm{P}}_{\mathrm{B}}\left(\mathrm{t}\right)\right)+\left(\sum {\int }_{{\mathrm{t}}_{2,\mathrm{k}}}^{{\mathrm{t}}_{\mathrm{C},\mathrm{k}}}{\mathrm{P}}_{\mathrm{C}}\left(\mathrm{t}\right)\right)+\left(\sum {\int }_{{\mathrm{t}}_{3,\mathrm{l}}}^{{\mathrm{t}}_{\mathrm{D},\mathrm{l}}}{\mathrm{P}}_{\mathrm{D}}\left(\mathrm{t}\right)\right)+\left(\sum {\int }_{{\mathrm{t}}_{4,\mathrm{m}}}^{{\mathrm{t}}_{\mathrm{E},\mathrm{m}}}{\mathrm{P}}_{\mathrm{E}}\left(\mathrm{t}\right)\right) \\ +\left(\sum {\int }_{{\mathrm{t}}_{5,\mathrm{n}}}^{{\mathrm{t}}_{\mathrm{F},\mathrm{n}}}{\mathrm{P}}_{\mathrm{F}}\left(\mathrm{t}\right)\right)+\left(\sum {\int }_{{\mathrm{t}}_{6,\mathrm{o}}}^{{\mathrm{t}}_{\mathrm{G},\mathrm{o}}}{\mathrm{P}}_{\mathrm{G}}\left(\mathrm{t}\right)\right) \end{gathered}$$

(2)

with P_A to P_G as the average power demand during the operating states A to G within the respective time intervals [t_0,i, t_A,i], [t_1,j, t_B,j], [t_2,k, t_C,k], [_t3,l, t_D,l], [t_4,m, t_E,m], [t_5,n, t_F,n], and [t_6,o, t_G,o].

The CO2PE!-methodology [6] was extended with the input model of [17]. The authors defined three energy demand shares for operating state A, which aggregate the energy demand of defined consumer groups. E_Basic sums the energy demand of the main switch and control system, lighting, hydraulics, cooling, ventilation, and other auxiliary units, E_Ready the energy demand of the suction, axis drive, chip conveyor, spindle, and tool, and E_Cutting the energy demand for the tool, cutting fluid supply and material removal. For transferring this logic to other operating conditions, the time component has to be disregarded, since it is specified by the individual machine tool programming. Therefore, the three average power demand states P_Basic, P_Ready, and P_Cutting are defined.

Table 3. Allocation of consumer groups to operating states and PKV. (PKV: Power Key Values).

				PKV Per Consumer Group POperating State–Consumer Group 1
Operating State	PKV Per Operating State PA to G 2			Main Switch and Control System	Lighting	Hydraulics	Cooling	Ventilation	Other Auxiliary Units	Suction	Axis Drives	Chip Conveyor	Spindle	Tool	Cutting Fluid Supply	Material Removal
A	PA			PA-MC	PA-L	PA-H	PA-C	PA-V	PA-O	PA-S	PA-AD	PA-CC	PA-Sp	PA-T	PA-CF	PA-M
	PBasic	PReady	PCutting
B	PB			PB-MC	PB-L	PB-H	PB-C	PB-V	PB-O	PB-S	PB-AD	PB-CC	PB-Sp	PB-T	PB-CF	-
C	PC			PC-MC	PC-L	PC-H	PC-C	PC-V	PC-O	PC-S	PC-AD	PC-CC	PC-Sp	PC-T	PC-CF	-
D	PD			PD-MC	PD-L	PD-H	PD-C	PD-V	PD-O	PD-S	-	PD-CC	-	PD-T	PD-CF	-
E	PE			PE-MC	-	-	-	-	-	-	-	-	-	-	-	-
F	PF			-	-	-	-	-	-	-	-	-	-	-	-	-
G	PG			PG-MC	PG-L	PG-H	PG-C	PG-V	PG-O	PG-S	PG-AD	PG-CC	PG-Sp	PG-T	PG-CF	-
Classification 3				c	c	c	c	c	c	c	c	c	v	v	c	v

¹ Active (bold), non-active (-) or active depending on the machine (italic writing) consumer groups per operating state according to [52]. The activity must be checked individually for each machine. ² P_Basic, P_Ready, and P_Cutting according to the logic of [17]. ³ Allocation to constant “c” and variable “v” consumer groups. It should be noted that this allocation may be different for different machine tool types.

shows the adapted allocation of the consumption shares according to the logic of [17] and the consumer groups (cf. Section 3.1) per operating state (cf. Section 3.2). According to [52], only the main switch and the machine control system appear as consumers within the operating state E. These consumers are implicitly considered in the approach of [17]. Therefore, the average power demand P_E must be introduced. While P_E is a power share of P_Basic for the operating states A, C, and D, P_E is used as a constant partial power share to map the power demand during E. The average power demand of the operating state B, P_B, as a separate mean power demand was introduced despite the proposition made in [17]. Although P_B aggregates the power demand of the same consumer groups as P_C, the average power demand of the two states is not the same and cannot be mapped with the power demand shares P_Basic and P_Ready. Due to the variable consumer groups, the power demand share P_Ready cannot be used to map the average power demand of the operating state C. Machine tools can also run through specific operating states due to individual machine tool programming. Therefore, P_G summarizes the average power demand of the operating state G.

From the allocation presented in Table 3, the mean power values within a specific time period per consumer group and operating state can be derived. The constant mean power values, once determined, are the characteristics of a specific machining tool, independent of the workpiece. Consequently, they can be used as calculation factors for the energy demand based on the operating time in [s] of each operating state. The variable mean power values are dependent on the workpiece characteristics and must be derived for each machining process individually. To denote the significance of the mean power values in EEMA, these parameters are called the PKV.

Considering the input model of [17] and the requirements for modeling the PKV per operating state (Table 3), the TED_M can be calculated either as:

$$\begin{array}{ll}{\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{M}}={\displaystyle \sum _{\mathrm{i}=0}}({\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}& +{\mathrm{P}}_{\mathrm{R}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{y}}+{\mathrm{P}}_{\mathrm{C}\mathrm{u}\mathrm{t}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{g}}){\mathrm{t}}_{\mathrm{A},\mathrm{i}}+{\displaystyle \sum _{\mathrm{j}=0}}{\mathrm{P}}_{\mathrm{B}}{\mathrm{t}}_{\mathrm{B},\mathrm{j}}+{\displaystyle \sum _{\mathrm{k}=0}}{({(\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}+(\mathrm{P}}_{\mathrm{C}}{-{\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}\left)\right)\mathrm{t}}_{\mathrm{C},\mathrm{k}}\\ & +{\displaystyle \sum _{\mathrm{l}=0}}{({(\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}+(\mathrm{P}}_{\mathrm{D}}{-{\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}\left)\right)\mathrm{t}}_{\mathrm{D},\mathrm{l}}+{\displaystyle \sum _{\mathrm{m}=0}}{\mathrm{P}}_{\mathrm{E}}{\mathrm{t}}_{\mathrm{E},\mathrm{m}}+{\displaystyle \sum _{\mathrm{n}=0}}{\mathrm{P}}_{\mathrm{F}}{\mathrm{t}}_{\mathrm{F},\mathrm{n}}+{\displaystyle \sum _{\mathrm{o}=0}}{\mathrm{P}}_{\mathrm{G}}{\mathrm{t}}_{\mathrm{G},\mathrm{o}}\\ & ={\displaystyle \sum _{\mathrm{i}=0}}{\left({\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}{\mathrm{t}}_{\mathrm{A}}+{\mathrm{P}}_{\mathrm{R}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{y}}{\mathrm{t}}_{\mathrm{r}}+{\mathrm{P}}_{\mathrm{a}\mathrm{i}\mathrm{r}}{\mathrm{t}}_{\mathrm{a}\mathrm{i}\mathrm{r}}+({\mathrm{P}}_{\mathrm{R}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{y}}+{\mathrm{P}}_{\mathrm{c}\mathrm{o}\mathrm{o}\mathrm{l}}+\mathrm{k}\dot{\mathrm{v}}{)\mathrm{t}}_{\mathrm{c}}\right)}_{\mathrm{i}}+{\displaystyle \sum _{\mathrm{j}=0}}{{\mathrm{P}}_{\mathrm{B}}\mathrm{t}}_{\mathrm{B},\mathrm{j}}\\ & +{\displaystyle \sum _{\mathrm{k}=0}}{({(\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}+(\mathrm{P}}_{\mathrm{C}}{-{\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}\left)\right)\mathrm{t}}_{\mathrm{C},\mathrm{k}}+{\displaystyle \sum _{\mathrm{l}=0}}{({(\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}+(\mathrm{P}}_{\mathrm{D}}{-{\mathrm{P}}_{\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{c}}\left)\right)\mathrm{t}}_{\mathrm{D},\mathrm{l}}+{\displaystyle \sum _{\mathrm{m}=0}}{\mathrm{P}}_{\mathrm{E}}{\mathrm{t}}_{\mathrm{E},\mathrm{m}}+{\displaystyle \sum _{\mathrm{n}=0}}{\mathrm{P}}_{\mathrm{F}}{\mathrm{t}}_{\mathrm{F},\mathrm{n}}\\ & +{\displaystyle \sum _{\mathrm{o}=0}}{\mathrm{P}}_{\mathrm{G}}{\mathrm{t}}_{\mathrm{G},\mathrm{o}}\end{array}$$

(3)

or with regard to the PKV per consumer group as:

$$\begin{array}{ll}{\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{M}}={\displaystyle {\displaystyle \sum _{\mathrm{i}=0}}}{\mathrm{P}}_{\mathrm{A}} {\mathrm{t}}_{\mathrm{A},\mathrm{i}}& +{\displaystyle \sum _{\mathrm{j}=0}}{\mathrm{P}}_{\mathrm{B}} {\mathrm{t}}_{\mathrm{B},\mathrm{j}}+{\displaystyle \sum _{\mathrm{k}=0}}{\mathrm{P}}_{\mathrm{C}} {\mathrm{t}}_{\mathrm{C},\mathrm{k}}+{\displaystyle \sum _{\mathrm{l}=0}}{\mathrm{P}}_{\mathrm{D}} {\mathrm{t}}_{\mathrm{D},\mathrm{l}}+{\displaystyle \sum _{\mathrm{m}=0}}{\mathrm{P}}_{\mathrm{E}}{\mathrm{t}}_{\mathrm{E},\mathrm{m}}+{\displaystyle \sum _{\mathrm{n}=0}}{\mathrm{P}}_{\mathrm{F}} {\mathrm{t}}_{\mathrm{F},\mathrm{n}}+{\displaystyle \sum _{\mathrm{o}=0}}{\mathrm{P}}_{\mathrm{G}} {\mathrm{t}}_{\mathrm{G},\mathrm{o}}\\ & ={\displaystyle \sum _{\mathrm{i}=0}}{(\mathrm{P}}_{\mathrm{A}-\mathrm{M}\mathrm{C}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{L}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{H}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{C}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{V}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{O}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{S}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{A}\mathrm{D}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{C}\mathrm{C}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{S}\mathrm{p}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{T}}\\ & +{\mathrm{P}}_{\mathrm{A}-\mathrm{C}\mathrm{F}}+{\mathrm{P}}_{\mathrm{A}-\mathrm{M}}){\mathrm{t}}_{\mathrm{A},\mathrm{i}}\\ & +{\displaystyle \sum _{\mathrm{j}=0}}\left({\mathrm{P}}_{\mathrm{B}-\mathrm{M}\mathrm{C}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{L}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{H}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{C}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{V}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{O}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{S}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{A}\mathrm{D}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{C}\mathrm{C}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{S}\mathrm{p}}+{\mathrm{P}}_{\mathrm{B}-\mathrm{T}}\right)\\ & +{\mathrm{P}}_{\mathrm{B}-\mathrm{C}\mathrm{F}} {\mathrm{t}}_{\mathrm{B},\mathrm{j}}\\ & +{\displaystyle \sum _{\mathrm{k}=0}}({\mathrm{P}}_{\mathrm{C}-\mathrm{M}\mathrm{C}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{L}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{H}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{C}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{V}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{O}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{S}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{A}\mathrm{D}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{C}\mathrm{C}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{S}\mathrm{p}}+{\mathrm{P}}_{\mathrm{C}-\mathrm{T}}\\ & +{\mathrm{P}}_{\mathrm{C}-\mathrm{C}\mathrm{F}}){\mathrm{t}}_{\mathrm{C},\mathrm{k}}\\ & +{\displaystyle \sum _{\mathrm{l}=0}}{(\mathrm{P}}_{\mathrm{D}-\mathrm{M}\mathrm{C}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{L}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{H}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{C}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{V}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{O}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{S}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{C}\mathrm{C}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{T}}+{\mathrm{P}}_{\mathrm{D}-\mathrm{C}\mathrm{F}}) {\mathrm{t}}_{\mathrm{D},\mathrm{l}}\\ & +{\displaystyle \sum _{\mathrm{m}=0}}{\mathrm{P}}_{\mathrm{E}-\mathrm{M}\mathrm{C}} {\mathrm{t}}_{\mathrm{E},\mathrm{m}}\\ & +{\displaystyle \sum _{\mathrm{o}=0}}{(\mathrm{P}}_{\mathrm{G}-\mathrm{M}\mathrm{C}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{L}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{H}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{C}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{V}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{O}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{S}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{A}\mathrm{D}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{C}\mathrm{C}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{S}\mathrm{p}}+{\mathrm{P}}_{\mathrm{G}-\mathrm{T}}\\ & +{\mathrm{P}}_{\mathrm{G}-\mathrm{C}\mathrm{F}}){\mathrm{t}}_{\mathrm{G},\mathrm{o}}\end{array}$$

(4)

with P_{Operating State – Consumer Group} as the PKV per consumer group within the operating states A to G in [W] and the respective periods of the operating state t_A,I, t_B,j, t_C,k, t_D,l, t_E,m, t_F,n and t_G,o in [s]. The number of times an operating state runs is represented by the indices i to o for the specific operating states A to G. For the description of the other parameters, please refer to the explanation of Equations (1) and (2).

TED_M considers the proportional energy demand of all operating states and consumer groups of the machine tool within a defined operating cycle. How often the operating state A runs or how many products have been processed within this period is represented by z. The TED of a machining process TED_P can be expressed as follows:

$${\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{P}}=\frac{1}{\mathrm{z}}\left(\mathrm{T}{\mathrm{E}\mathrm{D}}_{\mathrm{M}}\right)$$

(5)

3.5. Validation of EEMA and Data Acquisition Proposal

To validate EEMA, the TED_M is calculated from the PKV and results are compared with the direct measurement of the TED at the machine level covering the real energy demand in its entirety. From this comparison, possible neglect of energy demand shares due to missing components or losses becomes obvious. Primary data were gathered on the three-axis, vertical pick-up turning machine EMAG VLC 100Y (EMAG GmbH & Co. KG, manufactured 2014) within the operating states A, C, and E (operating cycle 3 h:28 min:7 s). According to [18], the energy demand of air cutting is only 2% to 3% lower than the energy demand of actual material cutting. Therefore, the cutting operation runs in air cut to avoid material influences. Raw data were obtained in a temporal resolution of one second via the machine’s internal Programmable Logic Controller (PLC) and via mobile measuring equipment (UMG604-EP, Janitza electronics GmbH, Lahnau, Germany).

Table A5. Allocation of consumer groups to operating states and power demand determination.

Consumer Group	Aggregate	Power Demand Determination
Main switch & control	Main switch and control system	Calculation
Lighting	Lighting	Calculation
Hydraulics	Hydraulic pump	Measurement
Cooling	Cooling	Calculation
Ventilation	Ventilation	Calculation
Other auxiliary units	Conveyor belt
Lubrication	PLC
Measurement
Suction	Suction	Measurement
Axis drives	x-axis; y-axis; z-axis	PLC
Chip conveyor	Chip conveyor	Measurement
Spindle	Spindle	PLC
Tool	Tool, turret	PLC

shows the allocation of the machine-specific aggregates to the consumer groups and the applied power demand determination. Primary data were obtained for the consumer groups’ hydraulics, other auxiliary units, suction, axis drives, chip conveyor, spindle, tool, and cutting fluid supply. For the consumer groups’ main switch and control system, lighting, cooling, and ventilation, there was only one aggregated value (

Table A6. Calculation of the PKV of EMAG VLC 100Y (EMAG GmbH & Co. KG, Salach, Germany; air cut).

EMAG VLC 100Y (EMAG GmbH & Co. KG, 2014), Specific Machining Process—Air Cut
Operating State	Measuring Time Total [hh:mm:ss]	Number of Cycles	PKV, Consumer Groups AM [W]									PKV, Operating State AM [W]
			Main Switch & Control, Lighting, Cooling, Ventilation	Hydraulics	Other Auxiliary Units	Suction	Axis Drives	Chip Conveyor	Spindle	Tool	Cutting Fluid Supply
AAir	02:57:56	13	745.7	349.4	2.1	93.9	184.1	77.8	3.6	100.0	872.3	2428.7
C	00:09:56	4	653.3	317.3	0.0	0.0	3.8	0.0	0.0	0.0	0.0	974.4
E	00:30:15	3	589.2	0.6	0.0	0.0	0.7	0.0	0.0	0.0	0.0	590.5
Classification 1			c	c	c	c	c	c	v	v	c

¹ Allocation to constant “c” and variable “v” consumer groups.

). The primary data were transformed into the PKV and calculated the TED_M (cf. Equation (4)). For the actual validation of the EEMA, the calculated values (TED_M, PKV per operating state) were compared with the respective measured primary data and the Normalized Root Mean Square Error, NRMSE was determined (Appendix E, Equation (7)) (cf. [55]). While for operating state A the NRMSE was 0.010, it was 0.029, and 0.004 for operating states C and E (

Table A7. NRMSE (EMAG VLC 100Y, EMAG GmbH & Co. KG; specific machining process—air cut).

Operating State	Energy Demand Calculated [kWh]	Energy Demand Measured [kWh]	n	ymax [W]	ymin [W]	NRMSE
AAir	7.2	7.5	10,680	11,309.2	−558.5	0.010
C	0.2	0.2	600	1225.4	945.8	0.029
E	0.3	0.3	1818	888.7	579.8	0.004
TED	7.7	8.0	13,098	11,309.2	−558.5	0.008

). The NRMSE of the TED_M regarding the measured TED was 0.008 (Table A7). With this assessment of the accuracy of the modeling approach under the conditions presented, the validity of the EEMA was presented and, thus, confirmed the suitability of the disaggregated calculation method for determining the energy demand of machining processes.

To easily implement the model in practice, a procedure for companies was proposed to quickly obtain the data they needed for EEMA. To calculate the TED, the PKV of the relevant consumer groups were derived from primary data on the machine tool. In the context of digitization, it is possible to perform continuous sensor-based recording and automated analysis of the power demand of the machine tool or specific consumer groups. For the reason of condition monitoring, among others, this feature is already partly applied in modern machine tools [55]. If continuous energy monitoring is not carried out, a temporary power measurement at the consumer group level is recommended for practicability. With mobile measuring equipment, this is easily feasible even if a company has a low level of digitization.

In planning the power measurement campaign, it is necessary to define

• Which operating states run on the machine tool;
• Which consumer groups are installed;
• How the specific power demand can be assessed.

Based on this, a measurement plan has to be developed and executed. From the primary data, the PKV, the amount, and the respective periods of the single operating states that had been run can be derived. The PKV were used to calculate the TED_M or the TED_P (Equations (4) and (5)), depending on the scope of the underlying investigation.

If the interest is in the comparison or benchmarking of machining processes and not in the manufacturing of an individual workpiece, it is recommended to measure the power demand of the machine tool in the standard machine setting during air cutting to avoid the influences of material types.

The relevant information should be documented as presented in

Table 4. Documentation of the relevant information for EEMA (PKV: Power Key Values).

Machine Tool XY (Producer, Year of Manufacture), Machining Process YZ
Operating State	Measuring Time Total [hh:mm:ss]	Number of Cycles	PKV Per Consumer Group 2													PKV Per Operating State
			Main Switch and Control System	Lighting	Hydraulics	Cooling	Ventilation	Other Auxiliary Units	Suction	Axis Drives	Chip Conveyor	Spindle	Tool	Cutting Fluid Supply	Material Removal
Classification 1			c	c	c	c	c	c	c	c	c	v	v	c	v
A	tA	#A	PA-MC	PA-L	PA-H	PA-C	PA-V	PA-O	PA-S	PA-AD	PA-CC	PA-Sp	PA-T	PA-CF	PA-M	PA
B	tB	#B	PB-MC	PB-L	PB-H	PB-C	PB-V	PB-O	PB-S	PB-AD	PB-CC	PB-Sp	PB-T	PB-CF	-	PB
C	tC	#C	PC-MC	PC-L	PC-H	PC-C	PC-V	PC-O	PC-S	PC-AD	PC-CC	PC-Sp	PC-T	PC-CF	-	PC
D	tD	#D	PD-MC	PD-L	PD-H	PD-C	PD-V	PD-O	PD-S	-	PD-CC	-	PD-T	PD-CF	-	PD
E	tE	#E	PE-MC	-	-	-	-	-	-	-	-	-	-	-	-	PE
F	tF	#F	-	-	-	-	-		-	-	-	-	-	-	-	PF
G	tG	#G	PG-MC	PG-L	PG-H	PG-C	PG-V	PG-O	PG-S	PG-AD	PG-CC	PG-Sp	PG-T	PG-CF	-	PG
AAir	tAir	#Air	PAir-MC	PAir-L	PAir-H	PAir-C	PAir-V	PAir-O	PAir-S	PAir-AD	PAir-CC	PAir-Sp	PAir-T	PAir-CF	-	PAir

¹ Allocation to constant “c” and variable “v” consumer groups. ² Relevant (bold), non-relevant (-) or relevant depending on the machine tool (italic writing) consumer groups per operating state. Individual relevance must be checked for each machine tool.

.

The added value of EEMA is the reusability of the constant PKV for calculating the TED of other machining processes on a specific machine tool. Only the variable PKV for a new machining process (P_A-Sp, P_A-T, P_A-M) need to be determined. These can be raised via the PLC (P_A-Sp, P_A-T) and obtained via calculation (P_A-M = k × $\dot{v}$, cf. [17]). Consequently, once a company has derived the PKV of a machining tool, the TED for all further production modes and machining process on this tool can be determined without further measurements.

3.6. LCI Datasets from EEMA

Following the principles for LCA databases according to [56], the generation of LCI datasets requires a mathematical relation of the primary data to the reference flow. This requirement is met with EEMA and the underlying PKV. Within the LCI, the input flow “electricity” for modeling the energy demand of a machining process is represented by the TED_P based on EEMA (Equations (4) and (5)).

For the derivation of LCI datasets, the reference flow was specified. EEMA provided the necessary connection to measurable parameters and the linear relation to the TED. It yielded, as a reference flow, the TED_P per second (cf. Figure 5). Following the idea of the CO2PE!-methodology [6], a reference second of a machining process was derived from EEMA (TED_P/s in [kWh/s], cf. Equation (6)), with Øt_A in [s] as the mean total duration of the operating state A within the considered operating cycle:

$${\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{P}/\mathrm{s}}=\frac{{\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{P}}}{\mathrm{Ø}{\mathrm{t}}_{\mathrm{A}}}=\frac{1}{\mathrm{z}\mathrm{\ast }\mathrm{Ø}{\mathrm{t}}_{\mathrm{A}}}\left({\mathrm{T}\mathrm{E}\mathrm{D}}_{\mathrm{M}}\right)$$

(6)

Ref. [56] underlined the requirement to report the most relevant supportive information within a dataset. However, the dataset analysis (Section 2.2, Appendix A) showed that only little meta-information about the machine tool was provided. For LCI datasets derived from EEMA, it is recommended to include the following supplementary information (cf. Table 4):

• Machine type, producer, year of manufacture, and type of machining process investigated;
• Existing operating states, at best including a short description of the process sequences;
• Existing consumer groups at the machine level;
• PKV of the identified consumer groups according to the respective operating states;
• PKV of the identified operating states;
• Measuring times and the number of iterations of the respective operating states.

Following the above-described framework and the PKV data acquisition described in Section 3.5, LCI datasets for machining tools can be easily generated and transparently documented in company databases for production sites or as the result of research projects on individual machining tools.

Furthermore, EEMA can support the compilation of generic LCI datasets. Usually, generic datasets aim to represent a certain production or market mix. As an example, datasets for electricity represent the supply mix in the grid within a certain reference area. However, for machining processes, a large variety of machining tools and a huge number of companies applying different machining processes exist, which makes it virtually impossible to derive such a representative market mix. With the help of EEMA, however, some important elements can be obtained to derive generic datasets.

The TED depicts constant and variable PKV. The share of the energy demand of the variable consumer groups is comparatively low (cf. [14]). To examine this further, an exemplary case study was used to examine the TED as a function of machine utilization, presented in Appendix F. The input data for our fictional case study was based on the measurements on the EMAG VLC 100Y (see Table A6). The key findings were as follows: The constant consumer groups had a significant influence on the TED_M. The energy demand share of the variable consumer groups in the TED_M increased with increasing machine utilization but was always below 5%. The increase in machine utilization also increased the TED_M per operating cycle, which confirmed the dominance of operating state A. At the same time, TED_P decreased as the production rate increased within the respective operating cycle.

Based on the generic PKV in conjunction with real operating periods, LCA practitioners can estimate the energy demand of machining processes reasonably well. The generic PKV can be derived from a representative number of real PKV of specific consumer groups for the different operating states. To minimize uncertainties concerning particularly energy-intensive consumer groups (e.g., cooling system, main drives), a higher amount of consumption data should be considered for calculating the generic PKV. Especially for product-oriented LCA studies, such a generic dataset would lead to more precise modeling and accounting.

However, a further simplification can be envisaged to derive a “synthetic machining process” for LCA databases. Due to the low influence of the variable energy demand share, this share can be set as a constant estimate for a generic dataset (e.g., 5% of the constant consumer groups in operating state A). From this simplification, it follows that it is not necessary to distinguish between different materials, since the type of material only affects the variable energy demand. To create a generic dataset, the following assumptions can then be made: Figure A1 shows the dependency of the energy demand share of operating state A in the TED on the machine utilization rate. This dependency can be used for a relation that is based on the constant PKV of the operating state only, by introducing correction factors from the machine utilization rate. As the only indispensable information, an estimate for a generic PKV has to be developed. This can be done with practical measurements and/or manufacturer information. The machine utilization rate can be provided at different levels in a dataset that can be used for sensitivity analysis.

4. Discussion

The comparison of the developed EEMA with the approach developed by the CO2PE!-Initiative [6] shows that, though they are closely related approaches developed for machining processes, they vary regarding the model granularity and the related data acquisition effort. The reusability of a single measurement campaign and, thus, transferability to other products are only given for the EEMA model.

The discussion of the elements of EEMA starts with a comparison of the developed EEMA core model as described in Section 3.4. Previous literature [9,26,27] provided models to determine the TED, but none of the models can depict the contribution of the constant and variable consumer groups, and, therefore, partial results or input parameters cannot be reused for follow-up investigations. However, EEMA is also based on several assumptions, resp. limitations. First, the definition of the generic consumer groups as well as the generic operating states was derived from a literature analysis. It is possible to adapt the chosen categorization in practice or for specific companies and their machinery. Additionally, it is essential to evaluate the quality of information provided by both the TED_M and the TED_P, depending on the machine utilization and the operating cycle being analyzed.

Despite these limiting factors, it was demonstrated with the validation of EEMA (Section 3.5) that the results were adequately accurate to be used as basis for LCI datasets.

Discussing the EEMA results from the fictive case study, this study confirmed previous findings regarding the importance of the constant consumer groups [9,11,12,14]. All studies stated that constant consumer groups have a significant influence on the TED. This article further specified that the share of energy demand of the variable consumer groups in the TED increased with increasing machine utilization but was always below 5% (see Section 3.6).

Section 3.6 further illustrates how EEMA results can be used in the future to generate generic LCI dataset for LCI databases. The analysis of present LCI datasets (Table A1 and Table A3) showed the broad scatter of values for the input flow “electricity”. Although it was shown that this scatter was partly due to deviating system boundaries, it can be assumed that a significant part was attributed to the arbitrary choices of a technical reference flow that did not depict the real dependencies of the TED. Consequently, even a rough but substantiated estimate of a generic dataset for machining processes based on EEMA will be a progress in view of a more realistic picture of the contribution of machining processes in product-oriented LCA. To adapt LCI datasets generated by EEMA to the system boundary for a complete assessment of machining processes as described in [36], other technical inputs can easily be added to the LCI datasets derived from EEMA (cf. Figure 5).

5. Conclusions

EEMA was derived from a thorough analysis of the literature on energy demand modeling approaches of machining processes and LCA requirements. The transferability of EEMA was based on the PKV, which indicate the mean power demand values per consumer group in relation to the operating states. PKV can be easily derived from measured primary data of a machine tool. By differentiating between constant and variable components, the constant PKV can easily be adopted for subsequent investigations on the machine tool. The variable PKV can be derived from machine parameters and do not necessarily have to be measured. Thus, EEMA provides the basis for calculating the TED for variable operating conditions and workpieces of real-world production based on a single measurement campaign. This allows companies to realistically and efficiently assess the TED and associated GHG emissions, which can be used for both machining process LCA and Scope 2 corporate Carbon Footprint assessments.

The definition of consumer groups and operating states is based on a literature analysis. In practice, it may be useful for organizations to define own consumer group or operating state categories for the specific machine tools used by the company. The EEMA can easily be adapted accordingly.

To derive LCI datasets, the approach of [6] and the requirements of [56] were followed. This enables companies and researchers to achieve clear documentation of datasets generated for individual machining tools.

In view of a possible derivation of generic LCI datasets, it was shown that operating state A had the determining energy demand share in the TED, especially at high machine utilization. Based on this, a simplified approach was outlined for a rough estimate of generic datasets from EEMA for a synthetic machining process to be integrated into conventional LCA databases.

EEMA can, furthermore, be used for benchmarking purposes and process-specific hotspot analyses of the most consumption-intensive consumer groups or operating states. The transparent comparison between machine tools or a previously defined reference scenario was easily possible due to the identification of the PKV and the structured documentation setup. Thus, the approach helps machine operators to detect optimization potentials within the process flow and supports machine tool manufacturers within the machine tool design phase.

The innovation of EEMA lies in its transferability of results from measured primary data based on the PKV, depicting the constant and variable consumer groups and the different operating states. Beyond assessing the energy demand of machining processes for a single product, our generic approach can be used to derive LCI datasets that guarantee a linear relation between the material removal and the TED. EEMA enables the reuse of the PKV for follow-up investigations on the specific machine tool and, thus, improves the efficiency of subsequent energy or environmental assessment investigations in terms of cost and time in practice.