COMBINING RENEWABLE ENERGY AND DESIGN-FOR-MANUFACTURING RESEARCH IN AN UNDERGRADUATE RESEARCH PROJECT

Transcription

1 COMBNNG RENEWABLE ENERGY AND DESGN-FOR-MANUFACTURNG RESEARCH N AN UNDERGRADUATE RESEARCH PROJECT John T. Tester 1 Abstract - Northern Arzona Unversty s (NAU) College of Engneerng and Technology (CET) s ntegratng the tradtonal topcs nvolved n desgn-for-manufacturng research and educaton wth the re-emergng feld of renewable energy. n partcular, student research projects were conducted for the optmzaton of njecton moldng process parameters for a small wnd turbne blade to produce the fewest manufacturng defects. The defect of nterest was warp the level of part dstorton predcted after the moldng process. Devaton from the desgned turbne blade geometry degrades the predcted energy output for the system, thus nsprng ths research. The research actvtes nvolved manufacturng smulaton and desgn-of-experments (DOE) n order to create and examne a mathematcal response surface. Presented heren s a background to the NAU Sustanable Energy Solutons research and educaton actvtes, the partcular problem formulaton for the undergraduate research, the educatonal approach necessary to support the undergraduate researchers, the mplementaton of the research actvtes, and the project outcomes. ndex Terms Desgn of Experments, DOE, Renewable Energy, DFM, njecton moldng, smulaton. NTRODUCTON Ths paper descrbes the research that combnes the two topc areas of renewable energy and desgn-formanufacturng nto one. They are ted together through an approach whch uses a combnaton of smulaton and desgn of experments, such that the ssue of reducng defects n manufactured products s ted to the geometry of wnd turbne blades. NAU/CET s a prmarly undergraduate nsttuton, wth an emphass on engneerng educaton. However, a majorty of the Mechancal Engneerng faculty are actve members f the Sustanable Energy Solutons group. Ths research group s focused on devsng sustanable energy solutons, especally as appled n the Amercan Southwest and on Natve Amercan lands. n pursut of ths goal researchers addresses educaton and projects n the areas: Research and Development Educaton Technology Transfer Project mplementaton Sustanable Product Realzaton The Sustanable Energy Solutons group s part of a larger NAU envronmental effort the Center for Sustanable Envronments. Ths larger organzaton brngs together regonal scentsts, educators, ndependent scholars, busness leaders, government agences, non-profts, students, and communty members to seek creatve solutons to envronmental problems. These problems nclude not only techncal ssues, but socetal and governmental ssues as well. The areas of Research, Sustanable Product Realzaton and Educaton are the man foc of ths actvty. n partcular, Product Realzaton s the term used to address the desgn and manufacture of renewable energy systems. Ths process s hstorcally addressed as manufacturng educaton and research areas; product realzaton s the prmary nterest of the faculty researcher. A small, flexble computaton and prototypng laboratory has been developed at NAT/CET for nvestgatng such sustanable product realzaton topcs. Ths faclty, termed RAPDLab, also serves as a laboratory for undergraduate and graduate desgn courses. Manufacturng s not the typcal genre for the general area of envronmental educaton, nor renewable energy systems n partcular. Coppola documents envronmental lteracy as beng best suted and typcally addressed n the lberal arts educaton envronment[1]. Wood even addresses a dearth of manufacturng research n renewable energy research (at least, for small-scale systems) [2]; yet that very area has the most pressng need for advancement, so that the captal expense of these systems may be reduced, thus allowng more ndvduals to obtan them. Gven the stmulatng nteracton between the varous academc dscplnes at NAU n envronmental ssues and renewable energy, the objectve for ths author was to create undergraduate research that combnes the techncal ssues n both renewable energy and manufacturng research. Even n engneerng publcatons, renewable energy research often emphaszes desgnng to mert, socal good, or a techncal performance measure generally ths measure s optmzed for effcency of energy output. Much of desgn and manufacturng research emphaszes desgnng to cost. Note that ths cost s often n terms of mnmzng scrap or defectve materals. The objectve of ths author was to combne the two nto a sngle research effort, wth the 1 John T. Tester, Assocate Professor, Northern Arzona Unversty, Department of Mechancal Engneerng, Box 15600, Flagstaff, AZ ; John.Tester@nau.edu S1E-10

2 purpose of ncludng undergraduate researchers n ths process. PROBLEM DEFNTON The most promnent renewable energy research n the NAU Mechancal Engneerng Department s n small wnd turbne systems. The aerodynamc performance effcency of these systems s a prmarly a functon of the propeller, or blade, desgn. The shape of the blade determnes ts aerodynamc lft as wnd passes across t. Ths lft drves the rotaton of the turbne hub, thus drvng the electrcal generator, producng electrcty. Wnd turbne blades, from a manufacturng pont of vew, can consume 10%-20% of the captal cost.[3] The usual practce for small wnd turbne blade manufacture s to ether produce the parts va mllng from sold wood stock or va a composte/sandwch constructon. Both of these alternatves are farly expensve, due to the large cycle tmes of the processes and nherent labor costs. However, these turbne blades are small enough that njecton moldng becomes a vable manufacturng technque. njecton moldng s a very economcal process, as the cycle tmes are very short, there s mnmal labor nvolved, and the materal (thermoplastc) s nexpensve. The njecton moldng process s the preferred method for one of the most prolferate small wnd turbne systems n the market, produced by Southwest Wndpower. njecton molded products have some dsadvantages for wnd turbne blades. The njecton moldng process, conssts of meltng a plastc charge, njectng t nto a hollow cavty of steel, allowng the mold to cool whle at a prescrbed temperature, and after a prescrbed tme (the cycle tme), ejectng the blade from the mold. As part of coolng, the lqud wll take up more volume than the cooled sold; the the sold part wll have less volume than the mold cavty from whch t s ejected. Ths shrnkage process tself can contrbute to a warpng problem, descrbed subsequently. njecton moldng s most common for parts of farly unform thckness. However, the wnd turbne blade has varyng thckness through at least ts cross-secton, due to the arfol shape requred as part of ts aerodynamc functon. Furthermore, the lqud cools and freezes as t travels through long, narrow cavtes. These events contrbute to uneven coolng n the plastc part; ths uneven coolng results n dsorganzed polymer molecules, whch results n dstorton or warpage of the fnal shape. As the shape of the part determnes the blade s power effcency, the qualty objectve of moldng the blades s to mnmze the dstorton. Dstorton can be so severe that the part s rejected as scrap the cost of that part s thus absorbed nto the remanng batch of accepted parts. The greater the ncdent of dstorton, the greater the scrap rate, and therefore the greater the ultmate cost of the accepted parts. There are many varable manufacturng parameters n the njecton moldng process that contrbute to warp of the parts. The njecton pressure of the molten plastc, the melted polymer temperature, and the mold temperature are wdely consdered as factors whch affect warp [4]. After the mold s flled wth lqud polymer, a pressure s appled to the cavty as the polymer cools to a sold; ths packng pressure s also consdered a sgnfcant factor n the dstorton mechansm [5]. Educatonal Approach There are several teachng methodologes n engneerng and technology desgn educaton; however, the two broadest cultures can be stated as ether nternal or socal [6]. nternal desgn methodology assumes that the process of desgn s ndependent of socal factors. Socal desgn educaton methodology, on the other hand, emphaszes the socety along wth the scence of the desgn process under consderaton. Both are of mert, but requre dfferent educatonal delvery methods. The nternal educaton methodology s better suted to preparaton for undergraduate research. These students are more comfortable n a procedural envronment ther sklls and knowledge are freshly establshed, but they are not yet adept at plannng and adaptng ther sklls to detaled problem-solvng. The student begns to understand the theory of both the techncal engneerng analyss ssues and the renewable energy topcs as he or she practces the theory. The NAU/CET educatonal envronment s more ntmate that larger research nsttutons; class szes rarely exceed 40 students. Students also are predomnately taught drectly by the nstructors nstead of through teachng assstants, vstng nstructors or adjuncts. On the other hand, CET s strvng mghtly to expand ts msson to nclude funded appled research, n lght of the ongong shrnkng budget contrbutons from the state of Arzona. The result of these two conflctng objectves personal professor/student educaton and emphass for appled research results n a straned currculum. Wth only 7 tenure/tenure-track faculty n the Mechancal Engneerng department, the electve course offerngs are lmted for the students. Most undergraduate electves are only offered every other year. The recent addton of masters-level graduate classes further decreases the undergraduate offerngs. Ths educatonal envronment s where the undergraduate research must be accomplshed. The problem statement nvolves manufacturng, desgn, and hgher-level statstcal analyss educaton n order to be successful. Yet the undergraduate researchers avalable do not have all of the tools necessary to fully accomplsh the tasks. The soluton s for the faculty member to structure the research tasks such that the student researcher can operate n a very structured task envronment. The student s provded the knowledge necessary to accomplsh tasks on a just-n-tme bass. The order n whch the student s presented ths JT nformaton as part of a research and learnng process s dscussed n the MPLEMENTATON secton. S1E-11

3 As the man thrust of the problem nvolves njecton moldng, the deal stuaton s to perform njecton moldng experments to produce data for analyss. However, the small sze of the college agan results n restrctons on the researcher. No njecton moldng machne s avalable for use; even f one were to become avalable, the current CET buldng would have nether the avalable floor space nor the faclty nfrastructure to sustan njecton moldng operatons. The alternatve s to perform the experments va computer smulaton. The smulaton software of choce s MoldFlow TM. Ths software s provded as part of the larger sute of CAD/CAM tools nsde -deas TM (an EDS product). The faculty researcher obtaned the software through grants, and made the software readly avalable at NAU/CET and n partcular n the faculty researcher s RAPDLab faclty. The njecton moldng smulaton software smulates njecton moldng under a varety of process parameter settngs. Thus, by settng the four parameters of nterest at vared values, and the rest at nomnal values, the research can be accomplshed. The next step n the research defnton was defnng how these experments was to be conducted. An ordered and controlled method of expermentaton was requred. Desgn of Experments Methodology The research has the mmedate techncal objectve of optmzng the manufacturng process for the shape at hand. There are methods for creatng theoretcal models of the problem; however, nearly all njecton moldng research nvolvng theoretcal formulaton s targeted at graduatelevel partcpants. However, f the manufacturng process n queston s well-defned, a useful alternatve to the optmzaton problem s by usng a desgn of experments (DOE) approach. The lterature search, combned wth knowledge provded from the faculty research, reduced the number of njecton moldng parameters to four; ths process of elmnaton s termed factor screenng. [7] These parameters are dentfed as: njecton pressure ( X 1 ) Packng pressure ( X 2 ) Mold temperature ( X 3 ) Melt temperature ( X 4 ) Optmzaton requres an output measure. Though the concept of part dstorton s clear, the measure must be specfc. A typcal process of measurement for qualty control of plastc products s by measurng a key geometry of a part after moldng. Ths measurement s compared to the deal measurement from the part desgn. The assumpton of ths measurement process s that the sngle datum measurement wll be representatve of the overall dstorton of the product, allowng for an acceptance/rejecton crtera for the ndvdual parts. n the smulatons, ths sngle-pont value for a smulaton j s termed the tp deflecton measurement ( TD j ). However, for proprety reasons, the exact sngle-pont measurement actvty cannot be revealed n ths paper. For smlar propretary reasons, the actual blade geometry s not revealed n ths paper; nstead, a generc blade desgn s used for llustraton. The overall dstorton s termed warpage. The measure of warpage was ntally left for the student to determne. Ths allowed for a level of student partcpaton that was to encourage the student researcher to have some ntellectual mpact n the research, nstead of beng left to merely run smulatons and collect data. Ultmately, wth the faculty mentor s assstance, the student correctly developed a warpage measure. t was based upon the fnte element method (FEM) mesh output. Each element had a specfc locaton, R r, n space before moldng (as part of desgn) F and after moldng, R r, (as a part of njecton moldng smulaton). The dfference between the two postons s collected for each element, multpled by each element s ntal area, A, summed, and the sum dvded by the total model area: AW m r R = 1 j = m = 1 r R Ths measure allowed an average warpage, AW j, for each smulaton experment j n terms of lnear unts. The average warpage s therefore n the same unts as that of the tp deflecton measurement at a specfc locaton. The results of multple smulatons can be analyzed for the two measure outputs, then compared to see f the tp deflecton qualty control used n practce would ratonally correlate wth the overall warpage of the blade. Once the DOE factors and the separate output measures were establshed, the research moved to determnng the DOE desgn. Based upon the faculty researcher s experence, a response surface for njecton moldng s typcally not a frst-order ssue n ths stuaton. Therefore, a second-order model was desred. The most thorough DOE desgn s a full factoral desgn, where all of the desgn space levels are examned. However, f all combnatons of the four desgn factors were run, wth each factor havng 3 levels, then the resultng analyss would requre a total of 3 4 = 81 smulaton runs. As most of these ndvdual runs take more that 24 hours to run on the avalable workstaton, such a DOE desgn was mpractcal. Another second-order, effcent, expermental desgn approach, whch results n the least bas and the best orthogonal error ft, s a central composte desgn (CCD). t allows for all combnatons of two-level factor runs, plus a center run (all factors at mdrange values) and then runs on each axs three of the factors are held at md-level values and the forth factor at an extreme-valued poston. CCD for four factors results n a S1E-12 A F A

4 requred set of 25 separate smulatons, whch s a reasonable number for ths problem. MPLEMENTATON Gven the above condtons, the student must accomplsh the followng tasks n order to optmze the njecton moldng process smulaton for a partcular blade desgn: 1. Research the lterature on a. Small wnd turbne geometres b. njecton moldng dstorton 2. Create CAD geometres 3. Run the moldng smulatons. 4. Collect data and statstcally analyze the response surface equaton. Tasks 1 through 4 can be vsually represented by the correspondng vsual output n Fgures 1, 2, and 3. These hghly techncal tasks requred a collecton of knowledge for both the faculty and student researchers. The faculty researcher had the knowledge, but the student researchers dd not. f all the knowledge was necessary for the students to have at the start of the research, then ths project would have faled, as the students dd not have the requste courses nor experence to conduct the experments. However, as seen earler, the tasks are essentally sequental n nature; thus, the delvery of the nformaton could be accomplshed n a JT fashon to match the tasks order. FGURE 1 OBTAN GENERC BLADE DMENSONS / CREATE CAD GEOMETRES FGURE 3 DSTORTON PLOTTED OUTPUT (ONE EXAMPLE OF 25 RUNS). The knowledge delvery method, combned wth the research tasks, are schematcally llustrated n Fgure 5. Each research task s dscussed n context of the knowledge delvery. Manufacturng Research The students search the publcatons for research on njecton moldng, specfc to part dstorton as part of the manufacturng process. The knowledge they requred n order to accomplsh ths task s partally obtaned from the ntroducton to Manufacturng Processes (ME 467), an electve course for junors and senors. The nstructor also conducts ndvdual, roundtable dscussons wth the students to dscuss the dfferences between academc research and practcal njecton moldng engneerng ssues. The end of ths task stage culmnates n the factor screenng process, though t should be noted that the student researchers need not understand fully the DOE process at ths tme. The lterature search also ncluded wnd turbne blade desgn characterstcs. The desgn geometry affects the njecton moldng process. The objectve s to create a blade geometry that s moderately generc for turbne systems under 1 klowatt. Such nformaton s lmted n academc crcles; thus the student researchers are encouraged to search ndustry and vendor publcatons as well. Snce the ultmate am for ths subtask s merely an assessment of dmensons, the lessenng of academc rgor s consdered acceptable. FGURE 2 MOLDNG MODELNG. S1E-13

5 FGURE 4 STATSTCALLY ANALYZE RESULTS N MATLAB. GRAPHCALLY, THREE FACTORS ARE PLOTTED WTH A FOURTH HELD CONSTANT (N THS CASE, NJECTON PRESSURE). FGURE 5 KNOWLEDGE DELVERY METHOD, NTEGRATED WTH RESEARCH TASKS. CAD Modelng Ths task requrements nclude modelng the blade geometry as well as programmng certan modelng tasks for subsequent smulaton automaton and data post-processng. All CET freshman are requred to pass a CAD fundamentals course. However, the researchers need advanced CAD sklls n order to be successful n ths task. The Advanced CAD/CAM (ME 351) course s a junor/senor electve whch addresses advanced modelng technques. t also touches on structural FEM. Though the theory of structural FEM s not drectly applcable to njecton moldng FEM, the CAD technques of model constructon are smlar. The faculty researcher has roundtable dscussons wth students on ths ssue, as well as presentng the students wth CAD automaton technques that were not presented as part of the ME 351 course. CET has several courses n dfferent dscplne areas that allow for students to engage n cross-dscplnary educaton. One example of ths concept s the MatLab/C Programmng (EE 222) course n the Electrcal Engneerng (EE) department. Though an EE delvered course, the ME students are requred to take ths course as part of ther plan of study. Ths course ntroduces the student researchers to command-lne programmng (C language) as well as nteractve programmng technques usng MatLab TM. Both are useful n processng data as t s transferred from CAD models nto smulaton scenaros, as well as post-processng the njecton moldng smulaton data. Smulaton n ths task, the students create and run the smulaton s as part of the DOE schedule. Though preferred, t s not requred that the students fully understand all of the DOE constructon methods at ths stage. The research was accomplshed n ths fashon; therefore, the faculty researcher conducted perodc roundtables n order to construct the smulaton schedule for the students as well as to check the output of the smulaton for errors. The ME 351 course s of value n ths task; thus, t could be delvered to the students at ths pont n tme. f that were the case, the faculty researcher must take a more actve hand n delverng some of the earler CAD knowledge to the students on an ndvdual bass. Smulaton usng MoldFlow s not taught n any context at NAU/CET; even the faculty researcher was only knowledgeable about the basc smulaton technques of ths tool. Thus, all researchers traned themselves n detals of smulaton wth MoldFlow usng the vendor-suppled, onlne tutorals. n an deal stuaton, the faculty researcher would send the student researchers to on-ste short courses for tranng. However, fundng was lmted and not avalable for ths approach. Process Results The smulaton runs are completed and the data s processed to determne the values of the output measures TD j and AW j for each smulaton j. These results for each X j are processed n order to ft two second-order, four-dmensonal curves (or response surfaces): One n terms of the measure TD and the other measure for AW. Statstcal analyss was accomplshed to determne the level of mpact each of the S1E-14

6 factors had on the outputs, as well as ther nteractons n the second order. These analyses requred some basc understandng of DOE fundamentals. The student researchers can attan ths knowledge through the cross-dscplnary electve course, Semconductor Desgn and Control (EE 472). Ths course ntroduces theory and practce of expermental desgn and statstcal process control; computer modelng and smulaton. Ths ntroducton, n the context of statstcal process control, s especally approprate for the research dscussed. t would be best for the student researchers to take ths course at the start of the research tasks. However, the students plans of study are too packed wth other electves at the start of ths research n order to take ths course early. Also, the course cannot be taken at an earler tme n the students career, due to prerequste restrctons. For ths partcular research task, the students were unable to ft EE 472 n ther plan of study; consequently, the faculty researcher conducted roundtable nstructon and ad-hoc assgnments n order to mpart the necessary DOE knowledge to the students. EDUCATONAL/RESEARCH OUTCOMES The research tasks took place over the course of a year (summer, fall and sprng semesters). As seen from Fgure 5, the deal stuaton would be for the research students to enroll n the requred courses before or durng the research tmeframe. The requste course represents 18 hours of semester credt. t was unlkely that any sngle student would have these courses; usng two students allowed for a range of educatonal experence appled to the project; nevertheless, both students together stll dd not have all the desred courses. Nevertheless, the followng educatonal objectves were accomplshed: 1. Expose students to practcal applcatons of Desgnfor-Manufacturng engneerng on a renewable energy desgn ssue. 2. Enable students to assst n complex desgn problem formulatons and partcpate n the soluton process. 3. Gve the students experence n applyng advanced smulaton and analyss tools. stuaton, the tp deflecton was at a mnmum when the average warpage was mnmum. Ths concluson was not a statstcal certanty for all blade forms, but may be a gude for future nvestgatons. ACKNOWLEDGMENT Ths research was accomplshed n part through NAU s Hooper Undergraduate Research Assstantshp and the NASA Undergraduate Research nternshp programs. The faculty researcher thanks the contrbutons of two undergraduate researchers, Ty Hargroder and Ben Metzger, who were recpents of these programs, respectvely. EDS s also acknowledged for provdng the educatonal grants of - deas TM software to ths nsttuton. Lastly, ntel Corporaton s acknowledged for provdng computatonal grants, allowng for computatonal machnes to be nstalled and used extensvely n the RAPDLab as part of ths effort. REFERENCES [1] Coppola, N. W., Greenng the Technologcal Currculum: A Model for Envronmental Lteracy, Journal of technology Studes, Vol. VXXV, No. 1, Summer-Fall, 1999, pp Clausen, P. D., Wood, D. H., Research and Development ssues for Small Wnd Turbnes, Renewable Energy, Vol. 16, 1999, pp T. D. Ashwll, Status of the DOE/Sanda Blade Manufacturng Program, 39th AAA Aerospace Scences Meetng and Exhbt, January 2001, AAA , pp Ehret, P., Davdoff, A., Jacque, F., Bung, H., Smulaton of the Complete njecton Cycle, Proceedngs of the 52nd Annual Techncal Conference ANTEC 94, Part 1, May , San Francsco, CA, USA, pp Publsher: Publ by Soc of Plastcs Engneers, Brookfeld, CT, USA, Bushko, Wt C., Stokes, Vjay K., Effects of dfferental mold temperatures on the warpage of njecton-molded parts, ASME, Materals Dvson (Publcaton) MD, v49, 1994, Advances n Computer-Aded Engneerng (CAE) of Polymer Processng, Proceedngs of the 1994 nternatonal Mechancal Engneerng Congress and Exposton, Nov , Chcago, L, USA, p de Vres, M. J., Desgn Methodology n Unversty Scence Technology and Scence (STS), Journal of technology Studes, Vol. XX, No. 2, Summer-Fall, 1997, pp Box, George E. P., Draper, Norman R., Emprcal model-buldng and response surfaces, New York : Wley, c1987. One research objectve was to determne the level of mpact the four factors had upon the output measure of average warpage. t was determned that one factor, njecton Pressure, had no mpact at all for the turbne blade shape examned. The other three factors Packng pressure, melt temperature and mold temperature were at ther maxmums n order to mnmze the average warp. A second research objectve was to determne the effectveness of the tp deflecton sngle-pont measurement aganst the average warp measurement. They were not hghly correlated; however, for the optmal (mnmum) average warpage S1E-15