Feature-based Energy Consumption Quantitation Strategy for Additive Manufacturing of Complex Parts


 Additive manufacturing (AM) has significant advantages over conventional manufacturing technologies, including freedom of design, mass customization, and the ability to produce complex structures. Because reducing manufacturing energy is an essential challenge for industrial sustainability and national economics, the trend toward AM motivates us to explore the energy consumption of AM to address energy efficiency. Existing energy consumption quantitation methods for AM parts require complex models, and the energy consumption characteristics of AM equipment muse be considered in the quantitative analysis process. Therefore, a feature-based energy consumption quantitation method for complex AM parts is proposed that uses a simple model and can be applied to different AM technologies. A feature segmentation method is first proposed to divide complex AM parts into typical AM features (AMFs). Then, the energy consumption model is developed for each AMF to quantify the energy consumption during fabrication of the entire part. Finally, the energy consumption characteristics of a typical mechanical part manufactured by three different kinds of AM processes—fuse deposition modeling (FDM), stereolithography (SLA), and selective laser melting (SLM)—are investigated using the proposed feature-based energy consumption quantitation method and measured in an experimental case study. The results show that the proposed method can effectively and quickly predict the energy consumption of AM part manufacturing. Moreover, the efficiency of different types of AM processes is compared and discussed to address applicable efficiency improvement methods. This method can predict the energy consumption of complex AM parts, and can be integrated into the AM three-dimensional software model, providing a reference for structural optimization of AM parts.


32
Additive manufacturing (AM) technology has developed rapidly in recent years, and is now a 33 widely used manufacturing process by which any complex digital model can be translated to an 34 entity quickly. The recent growth in AM has increased manufacturing industry return and 35 enhanced industrial competitiveness. China is paying more attention to the development of AM 36 technology. The report "Made in China 2025" states that it is important to accelerate the research 37 and development of AM technology and equipment. Three-dimensional (3D) printing is an 38 important AM technology. According to a research report on the market prospect and investment 39 uses high-energy beams (such as laser beams, electron beams, plasma, etc.), and the 48 manufacturing efficiency is low and the production cycle long because of layer-by-layer stacking. 49 High energy consumption increases the production cost of AM products and causes environmental 50 problems. For example, a large amount of fossil fuel combustion for power generation produces a 51 large amount of greenhouse gas (GHG) emissions (China's GHG emissions are up to 1.03 × 10 9 t 52 [6]), which increases the greenhouse effect. Therefore, the energy consumption and environmental 53 impact of AM have attracted a great deal of attention [7]. 54 At present, the energy consumption analysis methods for AM can be divided into two 55 categories. One is the direct energy consumption measurement of AM equipment in parts 56 processing. Laser AM is of particular concern for its high energy consumption, and it has been the 57 focus of many developments in this category. Baumers  The energy consumption dimension includes the primary and secondary energy consumption 155 of each manufacturing feature of the processed parts. The primary energy consumption used to 156 change the material properties. The secondary energy consumption included in this model is only 157 that directly related to part processing. The other energy consumed by the equipment has been 158 analyzed and discussed in previous studies [27,28], and is therefore not considered in this method. 159 post-processing. The printing time refers to the duration from the time of beginning workpiece 161 forming to the end of fabrication. The time consumed in the pre-processing and post-processing 162 stages of the parts is also considered in the entire time consumption evaluation. 163 The material consumption dimension is also an important factor in the environmental impact 164 of AM. It includes the direct material consumption for forming the part as well as the auxiliary 165 material consumed by printing supporting and adhesive structures. Moreover, the material 166 consumption is directly proportional to the energy consumption in the AM process. However, the 167 material consumption of AM parts was not the focus of this study; the proposed model directly 168 addresses only the energy consumption dimension. 169

Method description 170
AM parts are processed using layered stacking, and each manufacturing feature of the part 171 can be directly produced. The AM system can be divided into the following hierarchy, as shown in 172 Figure 2: the AM feature (AMF) layer, AM part layer, and AM equipment layer. All levels in the 173 system are related to each other through energy flow and material flow. An AMF is the basic 174 manufacturing unit of the system, and an AM part is composed of a series of AMFs. Therefore, the 175 energy consumption of each AM part can be obtained by analyzing the energy consumption of its 176 basic manufacturing features. Finally, the energy consumed by the material processing of the AM 177 equipment is also considered. 178 In the feature segmentation-based energy consumption analysis method, the AM parts are 179 divided into simple basic manufacturing features, AMF 1, AMF 2, … AMF n, on the basis of the 180 established, and the final energy consumption of the processed parts is determined by combining 182 the AMFs. In this strategy, a part is decomposed into an aggregation of AMFs to collapse its 183 geometric complexity for energy consumption calculation. Thus, the overall energy consumption 184 of a complex AM part depends on the energy consumption analysis of each AMF. Consequently   The geometric parameters, layering mode, and manufacturing direction of each AMF 219 modeled here are shown in Figure 3. According to the definition of the system boundary, the 220 energy consumed directly by processing the AMF is considered during the energy consumption 221 modeling process. Different processing paths and different AM methods will affect the energy 222 consumption and models. The following four processing paths are considered, as shown in Figure   The energy consumption of an AMF can be divided into two parts based on different AM 228 equipment features. One is the main energy consumption, that is, the energy consumption used to 229 change the material properties and form, such as the nozzle heating (in FDM), laser irradiation (in 230 SLA), or laser melting (in SLM). This part of the energy consumed in this process is influenced by 231 manufacturing parameters, and the processing parameters such as the layer thickness of the parts, 232 scanning speed, and scanning gap. The other part is the auxiliary energy consumption, that is, the 233 energy consumption of the actuator movement process, that is, the energy consumed by the 234 scanning action of the nozzle or laser, which is influenced by the processing path, the movement 235 of the plate, etc.. The energy consumption analysis for each basic AMF is as follows: 236 (1) Energy consumption model of the AMF hexahedron 237 Considering the geometric features of the hexahedron, and according to the slicing method 238 shown in Figure 3, each layer has the same shape and size. The main energy consumption required 239 for each layer of the AMF hexahedron, layer hex E , can be described as 240 where ahex and bhex are the bottom length and width of the AMF hexahedron, respectively; llay is 241 the layer thickness; and Pmat is the energy consumption required to change the properties and state 242 of the unit volume material. 243 14 The main energy consumption of the whole AMF hexahedron, main hex E , can be obtained by the 244 following: 245 where hhex is the height of the AMF hexahedron and k is the number of layers of AMF 246 hexahedrons. 247 The model of Pmat is different for different AM technologies. For FDM, Pmat can be obtained 248 using Eqs. (5) and (6). 249 where Pmat-h is the energy consumption of material heating, Pmat-m is the energy consumption of 250 material melting, c is the specific heat capacity (kJ/kg•°C), ρ is the mass of material per unit 251 volume(kg/m 3 ), ∆t is the temperature difference (°C), and q is the enthalpy of fusion (J/kg). 252 For crystalline materials or eutectic mixtures, which have a fixed temperature in the melting 253 stage (melting point), this part consists of two sub-components: heating energy and melting energy. 254 For non-crystalline materials, the temperature continues to rise during the melting process; 255 therefore, only Eq. (5) can be applied. 256 For SLA and SLM, in which a laser is used to change the material properties, Pmat-l can be 257 obtained by analyzing the laser energy consumption. 258 where ω is the material absorptivity, Plaser is the rated power of the laser, vlaser is the laser scan 259 Because the laser beam passes through the lens, it is a Gaussian beam, so the laser energy in 261 the spot follows a Gaussian distribution, and a parabolic shape is formed at the critical point of the 262 sintering cross section. Thus, the area can be written as 263 where hlas is the gap distance. 264 Therefore, Pmat can be also given as 265 las mat-l las las lay The auxiliary energy consumption of the AMFs is affected by the processing time and 266 movement speed of the actuator. Mathematically, the energy used for the auxiliary movement is 267 defined as follows: 268 where Paux is the power consumption of an auxiliary process, including the movement of the 269 worktable, scraper, and powder feeding device; t1 and t2 are the starting and ending times of a 270 single auxiliary action, respectively; and k is the number of auxiliary units. 271 Therefore, the total energy consumption of the AMF hexahedron, Ehex, is 272 (2) Energy consumption model of the AMF sphere 273 The AMF sphere is sliced as shown in Figure 3, and circles with different radii are obtained 274 for each sliced piece. The main energy consumption required for processing each layer of the 275 16 AMF sphere, layer sp E , can be given as follows: where rsp is the radius of the AMF sphere. 277 The main energy consumption of the whole AMF sphere, main sp E , can be obtained using the 278 following equation: 279 Therefore, the total energy consumption of AMF sphere, Ehex, is 280 (3) Energy consumption model of the AMF cylinder 281 Similar to the energy consumption analysis method of the AMF hexahedron, the AMF 282 cylinder is sliced as shown in where rcy is the radius of the AMF cylinder bottom. 286 The main energy consumption of the whole AMF cylinder, main cy E , can be given as 287 cy lay main 2 cy cy lay mat 1 where hcy is the height of AMF cylinder. 288 Therefore, the total energy consumption of AMF cylinder, Ehex, is 289   (4) Energy consumption model of the AMF circular cone 290 17 The AMF circular cone is sliced as shown in Figure 3, and the shape of each layer is a circle 291 with a gradually reduced radius. The main energy consumption required for processing each layer 292 of the AMF circular cone, layer sp E , can be expressed as follows: where rco is the bottom radius of the AMF circular cone and hco is the height of the AMF circular 294 cone. 295 Therefore, the total energy consumption required for processing the AMF circular cone, Eco, 296 is 297 (5) Energy consumption model of the AMF tetrahedron 298 The AMF tetrahedron is sliced as shown in Figure 3. The shape of each layer is an equilateral 299 triangle with gradually reduced side length. The main energy consumption required for processing 300 where h is the height of AMF, and ∆t is the interlayer auxiliary action time. 316 The general model of AMFs can be used to analyze and quantify the energy consumption of 317 all kinds of multi-feature AM parts. The basic AMFs can be obtained by the feature segmentation 318 method. In the process of feature segmentation, the divided basic features should be as simple as 319 possible to improve the efficiency and reliability of the energy consumption quantitation results. 320

Efficiency indicators of AM 321
Based on the proposed energy consumption quantitation strategy of AM and energy 322 efficiency analysis method for machine tools, the energy and production efficiency of AM Based on the energy consumption models developed in Section 2.3, and the energy 328 consumption stages of the AM process, including the pre-processing preparation stage, parts 329 shaping stage, post-processing stage, and machine waiting, the effective energy consumption of 330 the AM process is mainly used for the parts shaping process. Thus, the energy efficiency of 331 different AM machines can be expressed as 332 in which 333 where ηE is the energy efficiency of the AM machine, ET is the total energy consumption of the 334 whole AM process, and Ei is the energy consumption of different stages in the AM process, which 335 are pre-stage, forming stage, post-processing stage, and idle stage. 336

Time efficiency for the AM process 337
Time consumption of AM has a significant impact on inventory cost, production planning, 338 and scheduling in an AM factory. The time efficiency of AM can be defined as the ratio of the 339 effective forming time of a part to the total length of the AM process, which can be expressed as 340 where ηP is the time efficiency of the AM machine; tAMP is the forming processing time of the 341 parts; TT is the total time consumption of the whole AM process; Ti is the energy consumption of 342 different stages in the AM process, which are the pre-stage, forming stage, post-processing stage, 343 and idle stage. 344 3 Case study 345 To verify the accuracy and feasibility of the proposed feature-based energy consumption 346 quantitation strategy, a case study was conducted. The energy consumption models of typical 347 AMFs were applied to evaluate the energy consumption of the AM parts shown in Figure 5  The process parameters of the three types of AM processing method were determined to 356 ensure the forming quality of the processed parts. The main parameters for FDM, SLA, and SLM 357 are shown in Table 2. 358

Feature division of the manufactured AM part 360
According to the shape features of the AM parts and the proposed feature segmentation 361 method, the parts are divided into seven AMFs, as shown in Figure 6. The obtained AMF1 is a 362 cylindrical feature, AMF6 is a removed cylinder feature, and AMF3, AMF4, and AMF5 are 363 hexahedral features. The manufacturing features of AMF2 are approximated by hexahedral 364 manufacturing features. AMF7 is a quarter cylinder manufacturing feature, and four AMF7s form 365 a whole cylinder.

Theoretical energy consumption model of the investigated AM part 369
Based on the energy consumption analysis and modeling method of typical AMFs, and the 370 shape and size parameters of the processed AM parts in Table 1, the total energy consumption of 371 the investigated AM part, EAMP, is obtained based on each AMF in Figure 6, which can be 372 expressed as 373 For the FDM process, the auxiliary energy consumption is consumed by the process of plate 386 movement, maintaining the plate temperature and nozzle movement, and the energy consumption 387 of the auxiliary system of the LCD screen and LEDs, which can be described as follows: 388 where Pmot1 is the driven motor power of the plate, Phea is the power for maintaining the plate 391 temperature, Pmot2 is the driven motor power of the nozzle, Psys is the power of the machine 392 auxiliary system, and T is the part processing time. 393 For the SLA process, the auxiliary energy consumption for the parts mainly includes the 394 energy consumption of plate moving, scraper movement, and auxiliary system, which can be 395 calculated by the following equation: 396 where Pmot3 is the driven motor power of the scraper. 397 For the SLM process, the auxiliary energy consumption includes the energy consumption of 398 plate moving, scraper movement, power feeding movement, protective gas filling, cooling system, 399 where Pmot4 is the driven motor power of the powder-feeding device, Pcool is the power of the 401 cooling device, and ∆ti (i = 1-9) is the action time of the corresponding moving parts. 402

Energy consumption measurements of different printing modes 403
An AWS2103S Plus power measurement three-phase power analyzer was used to perform the 404 machine), UNIONTECH-RS3000 (SLA machine), and BLT-S210(SLM machine). This power 406 meter has an accuracy of 0.5% ± 0.15% of range error and ± 0.15% of sensor error, with a 0.
where Pt is the measured apparent power and cos φ is the power factor. In the experiment processes, three kinds of AM machines were used to process the parts 426 shown in Figure 5 under the process parameters listed in Table 2. The voltage probe was plugged 427 into the two-phase power inlet of the AM machines to measure the voltage value in real time. The 428 current clamp was plugged directly into the stripped supply power cable in the power socket exit. 429 Hence, the energy loss due to the electric transformer was taken into account for the electrical 430 energy study. The measured instant power data were saved and processed to the power meter 431 software on a computer. Figure 7 shows the electrical assembly diagram. For the SLA forming process of a given part, after the processing model is imported into the 452 AM machine, the processing begins with work platform preparation, as shown in Figure 9. In the 453 initialization stage, the workbench moves to the highest level of resin to prepare for the part 454 forming process. The energy consumption in this stage depends on the power of the workbench 455 29 motor and other auxiliary structures of the machine. After that, the part is processed in the next 456 stage, which includes the forming process of the supporting structure and part structure. Although 457 there is no cantilever or hollow structure in the designed part, a 3 mm mesh solid support is added 458 at the bottom of the designed model for easy removal of the part from the workbench and to 459 ensure the dimensional accuracy of the part. The electrical energy consumed in the part processing 460 stage is determined by the part volume, build direction, support design, number of layers, and 461 machine auxiliary energy consumption. After the part has been finished, the workbench moves up 462 slowly from the resin to the highest position, and then the machine waits. 463 For the SLM forming process of a given part, after the processing model is imported into the 464 AM machine, the SLM process begins in a pre-processing stage, as shown in Figure 10 Table 3. The deviation rate ∆Ej between the part processing energy obtained by the models and 491 experiments can be calculated using Eq.(43). A comparison of the results is shown in Figure 11. 492 The comparative results ( Figure 11) show good coincidence for different AM technologies 496 between the part processing energy consumption obtained from the proposed models and the 497 experiments. The deviation is within the range of 10%, which means that the model catches the 498 forming principle of the AM process. The energy consumption of the same part processed by 499 different AM methods is quite different. From the efficiency analysis, it can be seen that the 500 energy efficiency of the three different AM machines is approximately 80%, among which the 501 FDM machine has the highest energy efficiency in the parts processing process. However, the time 502 efficiency of the machines is relatively low: 76.55%, 73.85%, and 63.35% for FDM, SLA, and 503 SLM, respectively. The time efficiency of the SLM process was the lowest. 504

Deviation analysis 506
The comparison of the theoretical and experimental results verified that the developed 507 models are adequately accurate for the energy consumption prediction of AM parts manufacturing. 508 However, the process energy in all experimental runs was overestimated, which is a systematic 509 deviation related to the error of the auxiliary system energy consumption between the calculated 510 theoretical energy consumption and measured experimental energy consumption. The main reason 511 is that the power of each energy consumption unit in the auxiliary system fluctuates in the 512 processing process, and there is a gap between the actual power and its rated power value. The 513 error between the model and actual energy consumption increases gradually with the number of 514 energy consumption units in the AM system. The gap between the model and experience can be 515 further reduced by considering the power fluctuation of the auxiliary energy consumption units. 516 essential to simplify the complexity of the calculation, which may cause deviation from the true 518 energy consumption quantity. In these models, some AMFs with irregular shapes and structures 519 obtained after segmentation are simplified. For example, the manufacturing features of AMF2 are 520 simplified into hexahedral manufacturing features in model development, but it is unchanged in 521 the experimental process, which also causes a systematic deviation. In order to achieve more 522 accurate models, some correction coefficients could be used to decrease these deviations. 523 Finally, in the experimental energy consumption measurement, the electrical energy 524 consumption of the input terminal is measured directly by a power meter. Hence, the energy loss 525 due to the electric transformer was taken into account for the experimental electrical energy study. 526 In addition, owing to machine aging, reduced accuracy of the machine control system, the 527 scanning speed, and the power of the laser could also lead to deviation. 528

Power consumption feature analysis 529
From the forming stage of the FDM process, many disruptions can be observed during this 530 last stage. These oscillations are not generated by the stepper motors. Indeed, values of current and 531 voltage are the same throughout the forming stage because of the nature of the motors. 532 Oscillations are generated by the maintenance of the heat-bed plate. Figure 12(a) shows a 533 zoomed-in section of the forming stage. This physical phenomenon lasts only 1-2 s and is 534 repeated throughout the forming stage. 535 From the power consumption vs. time curve of the whole SLA process, it can be seen that the 536 forming power of the machine is basically stable within a certain power range, and there is no 537 34 large power fluctuation. A zoomed-in section of the SLA forming stage is shown in Figure 12(b). 538 The reason is that the laser power of the SLA machine is small, and a plate warm-up stage is not 539 needed. The energy consumed in the whole forming process is mainly determined by the auxiliary 540 system, stepper motor of each axis, and laser galvanometer motor. Hence, the power fluctuation is 541 not large in the SLA process. 542 For the power consumption vs. time curve of the SLM process, the power consumption in the 543 processing stage shows periodic fluctuations of different frequencies with large amplitude and 544 small amplitude, as shown in Figure 12 can be found that although the energy efficiency of AM machines is higher, there is still room for 557 improvement. Based on power consumption-time curves for FDM, SLA, and SLM machines and 558 power consumption feature analysis for different AM processes, the following measures can be 559 taken to improve the energy efficiency and production and processing time energy efficiency of 560 AM: 561 (1) Reduce auxiliary energy consumption of equipment. The reason for the relatively low 562 energy efficiency of the SLA and SLM machines is that their auxiliary energy consumption is 563 higher. Therefore, for different AM machines, there are different means to reduce the auxiliary 564 energy consumption. 565 For the FDM machine, select a reasonable heating plate and nozzle heating temperature, and 566 optimize the machine cooling system to reduce the energy consumption of the heating and cooling 567 processes, and reduce the waiting time of the machine. 568 For the SLA machine, the auxiliary energy consumption caused by too many auxiliary energy 569 consumption units can be improved by optimizing the equipment structure and simplifying the 570 auxiliary unit. 571 For SLM machines, the energy efficiency of the auxiliary equipment of the machine should 572 be optimized, such as improving the energy efficiency of the equipment cooling system and 573 reducing the waiting time of the machine. 574