DEVELOPMENT OF DIELECTRIC SENSOR TO MONITOR THE ENGINE LUBRICATING OIL DEGRADATION

S1061 DEVELOPMENT OF DIELECTRIC SENSOR TO MONITOR THE ENGINE LUBRICATING OIL DEGRADATION by Vasanthan BALASHANMUGAM * and Devaradjane GOBALAKICHENIN Department of Automobile Engineering, Anna University, M. I. T Cam pus, Chromepet, Chennai, Tamil Nadu, In dia Orig i nal sci en tific pa per DOI: 10.2298/TSCI16S4061B Introduction Pres ent day prac tice of fol low ing fixed sched ules of oil change in ter vals could re - sult in loss for the equip ment owner, as the oil is not uti lized up-to its max i mum use - ful life. Sim i larly, the ex tended use of en gine oil be yond max i mum use ful life is of high risk, which could lead ir re vers ible and cat a strophic dam ages to en gine parts. En gine oil con di tion in di cates the con di tion of en gine parts, in any ap pli ca tion. There fore, mon i tor ing the con di tion of the oil in real time is of par a mount im por - tance. Re search ers had es tab lished that the en gine oil deg ra da tion cor re lates with change in di elec tric prop erty of the en gine oil. The im por tant fac tor to re al ize the on-line real time mon i tor ing of the changes in di elec tric prop erty of the en gine oil is, the cost of di elec tric sen sor within af ford able limit for an op er a tor. Cur rent work aims at de vel op ing such a low cost af ford able di elec tric sen sor and engine oil sam - ples (SAE 15W40 grade) were col lected from du ra bil ity test en gines used in en gine test rig and on-road ve hi cles. These sam ples were tested for phys i cal and chem i cal prop er ties. Any changes in the prop er ties, of en gine oil mon i tored,in di cate that it un der goes deg ra da tion due to us age. A pro to type of ca pac i tive type sen sor was de - vel oped and val i dated with ref er ence flu ids. The di elec tric val ues mea sured us ing proto type sen sor in the used oil sam ples show a cor re la tion with change in phys i cal prop er ties. This trend and thresh olds of di elec tric pro vides ef fec tive plat form to mon i tor the en gine oil deg ra da tion. The sen sor could be cou pled to a suit able warn - ing de vice by in cor po rat ing spe cific al go rithms. Key words: en gine oil, deg ra da tion, di elec tric con stant En gine man u fac tur ers nor mally de fine the oil change in ter vals rather than spec i fy ing the max i mum use ful life of oil. The oil change pe riod ba si cally de rived based on ap pli ca tion re - lated con sid er ations such as dif fer ent op er at ing speeds and loads, op er at ing en vi ron ment, and dif fer ent duty cy cles. The light duty cy cle and se vere duty cy cle are the two ex treme con di tions un der which the same oil is used. The change pe riod, thus pre scribed does not in di cate the ac tual con di tion of the spe cific en gine oil. The change in chem i cal or phys i cal prop er ties of the used oil can be tested in reg u lar in ter vals to mon i tor the re main ing use ful life, in a lab o ra tory. This in - cludes mea sur ing deg ra da tion in chem i cal or phys i cal prop er ties, soot per cent age (par tic u larly in Die sel en gines), wear de bris and con tam i na tion due to fuel or wa ter di lu tion. This method is a te dious, ex tremely dif fi cult, ex pen sive, and not so prac ti cal pro cess as it in volves fre quent sam - * Corresponding author; e-mail: sriram_bvias@ya hoo.co.in

S1062 Balashanmugam, V., et al.: Development of Dielectric Sensor to Monitor the Engine... pling and test ing at lab o ra tory. Ma jor draw back of such pro ce dure is that it fails to pro vide the prior warn ing to the op er a tor about ex act end of use ful life, which is of fun da men tal need, if the oil has to be uti lized to its max i mum use ful life. There fore, mon i tor ing the con di tion of the oil, on ac tual real time op er a tion is there fore of par a mount im por tance to avoid dis card ing the oil much be fore its max i mum use full life as well as to avoid the risk of us ing de graded oil. Many re search ers re ported about ca pac i tance sen sors to mea sure the di elec tric con - stant of the en gine lu bri cat ing oil in or der to mon i tor the en gine oil deg ra da tion. The ca pac i tor sen sor has ca pa ble to mea sure the deg ra da tion de tec tion due to ox i da tion, wa ter con tam i na tion and wear par ti cle [1]. The grid ca pac i tive sen sor also used to com pute the di elec tric con stant of the lu bri cat ing oil con tam i na tion and found that the value of di elec tric con stant var ied from 6.5 to 10 pf in re la tion to the in put fre quency [2]. Agostin et al. used leaved-disc ca pac i tor to mea - sure the di elec tric con stant then cor re lated it with vis cos ity of the en gine lu bri cat ing oil [3]. The re searcher de signed a wire less send ing sys tem, which trans mits lu bri cat ing oil ca pac i tance in - for ma tion and en ergy be tween sen sor and reader for au to mo biles with a ca pac i tive sen sor [4]. Liu et al. con ducted ex per i men tal val i da tion on permittivity as a method of mea sur ing oil deg ra da tion and con firmed the change in trend be tween the permittivity and acid value, iron con tent, mois ture [5]. The ef fect of tem per a ture on di elec tric con stant of lu bri cat ing oil was stud ied by Tor rents and Areny [6]. They have showed that in the lower fre quency range, the tem per a ture co ef fi cient of the di elec tric con stant de pends on whether the oil is new or it has been used. Apart from pre vi ously mentioned sen sors, some com mer cially avail able sen sors are also ca pa ble of on line oil qual ity de tec tion by way of in ter pret ing lu bri cat ing oil di elec tric prop - erty but they are costly. In this re search work, an at tempt is made to de velop af ford able di elec tric sen sor to mon i tor the en gine oil con di tion. Engine oil degradation process As lubricants ages, they become less capable of delivering expected performance largely due to the pro gres sive dump of high amounts of sludge and in sol u ble com pounds into the oil sump. The pri mary driver for this prob lem is ox i da tion. Ox i da tion is a gen eral term used to de - scribe a com plex and se ries of chem i cal re ac tions, which dis turbs the chem i cal sta bil ity of the liq - uid and en cour ages for ma tion of new un wel come mo lec u lar spe cies within a lu bri cant sump. Ini - tially, oxidation was characterized as a chemical reaction involving oxygen. The definition has been ex panded to in clude any re ac tions in volv ing elec tron trans fer. Craft [7] de scribed the lu bri - cant oxidation as a three-stage process: Initiation, propagation, and termination [7]. Ini ti a tion stage in volved with the for ma tion of a free rad i cal, an atom or mol e cule frag - ment with one or more un paired elec trons. The big gest con trib u tor of free rad i cals is the ox y gen it self. Con tam i nants those are rich with ox y gen (air, wa ter) feeds ox y gen to the sys tem. Free rad - i cals are highly re ac tive and un-sta ble, quickly com bin ing with HC com po nents to form alkyl rad i cals and hydroperoxi-rad i cals. The prop a ga tion stage oc curs when hydroperoxi (per ox ide) rad i cals re act with the base oil or ad di tives to re gen er ate an alkyl-rad i cal (or gen er ate an al co hol and wa ter) and re start the cy cle. When high tem per a tures ex ist, the per ox ide rad i cals split and sus tain the chem i cal re ac tion. When wear de bris is pres ent, per ox ides may cat a lyt i cally split to sus tain the re ac tion, even at low tem per a tures. The prop a ga tion stage be comes autocatalytic, with the chem i cal re ac tions them selves pro vid ing the feedstock to start the next cy cle. The ter mi na tion stage oc curs when the des ig nated ox i da tion in hib i tor (an ti ox i dant) per forms its func tion. All the three stages of en gine oil deg ra da tion lead to for ma tion of acidic com po nents, which af fect the di elec tric con stant of lu bri cat ing oil as shown in fig. 1.

S1063 Fresh lubricating oil properties Ta ble 1 shows fresh oil prop - er ties of SAE 15W-40 grade lu - bri cat ing oil. The typ i cal value of ki ne matic vis cos ity at 100 C ranges from 14.5 to 15.7. The base num ber value is 9.6 ini tially and de creases as deg ra da tion takes place. When the lower value of base num ber equals acid con - tent, it shows end of oil life. The soot level be low 4% is de sir able. Figure 1. Degradation process of engine oil En gine oil sam pling and lab testing Ta ble 1. Fresh oil prop er ties Oil 1 Oil 2 En gine oil sam ples were col lected from Properties en gines from test bed and on road ve hi cles Values Values (die sel pow ered ve hi cles) at reg u lar in ter - Grade 15W-40 15W-40 vals. These sam ples were tested as per Density, [kgm 3 ] 881 875 ASTM pro ce dure and test re sults were com - Vis cos ity at 100 C, [cst] 15.7 14.5 pared with thresh old lim its. The anal y sis of col lected test sam ples aimed at prog nos tic of Vis cos ity index 131 135 the en gine oil brings out in ter est ing facts that Pour point, [ C] 30 27 the SAE 15W-40 grade of oil used in dif fer - Flash point, [ C] 204 230 ent en gines does not be have the same way. Sulphur, [wt.%] 0.32 Not traceable More over, the life of the oil does not de pend on a sin gle prop erty or fac tor un der con sid er - Base num ber 9.6 10.1 ation. In stead, any prop erty that over shoots Sulphated ash. [wt.%] 1 1.3 its limit first, de ter mines the end of life Phosphorous, [wt.%] 0.12 Not trace able though other prop er ties might be still within Zinc, [wt.%] 0.13 Not trace able their lim its. tab. 2 gives sum mary on lab test re sults of the sam ples col lected from En gine test bench. It is ob served that the use ful life of oil var ies in the sam ples de pend ing on var i ous fac tors. Ta ble 3 gives sum mary on lab test re sults of sam ples col lected from fleet ve hi cles. It is ob served that the max i mum use ful life of oil var ies in the sam ples de pend ing on the var i ous fac - tors. Such vari a tion in physio-chem i cal prop er ties of sam ples, col lected ex actly in same drain pe riod, de pends on fac tors like op er at ing tem per a ture, load cy cles, speed cy cles, wear ma - Ta ble 2. Oil anal y sis of sam ples from test bench en gines at reg u lar in ter vals En gine no. Test du ra tion [h] Sam pling fre quency [h] Con dition of oil Max life [h] Max life lim ited based on 1 350 50 De graded 250 Vis cos ity be yond 18.5 cst and soot be yond 4% 2 500 50 De graded 500 Sil ica reached max level of 40 ppm

S1064 Balashanmugam, V., et al.: Development of Dielectric Sensor to Monitor the Engine... Ta ble 3. Sum mary of oil anal y sis of sam ples from on road fleet of ve hi cles Sam ples col lected from Test du ra tion [km] Sam pling fre quency [km] Condition of oil Max life [km] Rea son for lim it ing max life Fleet 1 60000 30000 Good 60000 Oil con di tion checked up to 60000 only Fleet 2 60000 30000 De graded 60000 Fleet 3 60000 30000 Good 60000 Fleet 4 60000 20000 Good 60000 To tal base number reaches lower limit of 5 mg KOH/g Oil con di tion checked up to 6000 only te ri als, con di tion of pis ton rings/lin ers/valve seats/valve guides, in jec tor spray pat terns, in jec tor pres sure, air cleaner ef fi ciency, and so on. As such, the fixed change sched ule pre scribed in the ve hi cle does not truly re flect the en tire life, con sid er ing safe sit u a tion de pend ing on op er at ing pa ram e ters. Hence, it could be a risk in ei ther way to per mit the en gine to run for the pre scribed du ra tion of oil change pe riod, due to the fol low ing rea sons: the oil might have lost its life be fore the pre scribed pe riod (un der more se vere op er at ing conditions, which were not considered by the equipment manufacturer while validation), and the oil has been dis carded much be fore the end of use ful life. The de graded oil could lead to costly fail ures such as jour nal bear ing sei zure, pis ton ring sei zure, cam bush wear, etc. Development, validation and testing of dielectric sensor proto type The di elec tric sen sor has two co-ax ial cyl in ders, as elec trodes, which sep a rated by space or me dium and works on the prin ci ple of ca pac i tance. The me dium has cer tain di elec tric con stant. The di elec tric con stant is the mea sure of a ma te rial's in flu ence on the elec tric field. The net ca pac i tance will in crease or de crease de pend ing on the type of di elec tric ma te rial. Permittivity re lates to a ma te rial's abil ity to trans mit an elec tric field. In the ca pac i tors, an in creased permittivity al lows the same charge to be stored with a smaller elec tric field, lead ing to an in creased ca pac i tance. The value of ca pac i tance is mea sured by us - ing LCR me ter. The ca pac i tance can be stated in terms of the di elec tric con stant by Carey [8] and Baxter [9]. A co ax ial type sen - sor with in ner and outer elec trode is de vel - oped as shown in fig. 2. The di elec tric con - Figure 2. Protot type dielectric sensor stant was mea sured in ref er ence flu ids like to lu ene and co co nut oil. Ini tial er ror of Ta ble 4. Val i da tion re sult of sen sor 4-5% was ob served as shown in tab. 4. Fluid Mea sured [pf] Mea sured di elec tric Ref er ence di elec tric Er ror To lu ene 78.66 2.18 2.3 5% Co co nut oil 108.46 3.004 2.9 4% e r C C 0 where e r is the di elec tric con stant, C the ca pac i tance in far ads with di elec tric be - tween plates, and C 0 is the ca pac i tance in the ab sence of di elec tric.

S1065 Figure 3. Exploded view of modified sensor parts The er ror ob served was due to fringe & spac ing ef fects. A ground ter mi nal was in tro - duced to re duce fringe ef fect. A small ri fle hole of 2 mm di am e ter in in ner elec trode and lead through it was in tro duced to re duce the lead ef - fect. Di a met ric clear ance was re duced from 4mm to 2 mm be tween the elec trodes to re duce spac ing er ror. Ny lon bush ing in sleeves was in - tro duced to re duce stray ca pac i tance. The mod i - fi ca tions were done in or der to re duce er rors due to fringe ef fect, lead ef fect and spac ing ef fects [10]. With the mod i fied sen sor, the er ror was re - duced from 4 % to 1%. The mod i fied sen sor is shown in figs. 3 and 4. The ca pac i tance was mea sured for fresh lu - bri cat ing oil us ing a LCR me ter cou pled to the sen sor ground ter mi nal as shown in fig. 5. Ta ble 5 shows the vari a tions ob served with op ti mized proto type sen sor out put af ter im prove ment (1% max). Ta ble 5. Val i da tion re sult of mod i fied sensor Ref er ence fluid Di elec tric con stant (measured di elec tric) Di elec tric con stant (stan dard value) Error [%] To lu ene 2.319 2.3 1 Co co nut oil 2.934 2.9 1 Figure 4. Assembly view of modified sensor parts 1 inner electrode, 2 outer electrode, 3 top cover Nylon, 5 top sleeve, 6 bottom Bush Nylon, 7 bottom sleeve, 8 ground terminal, 9 negative terminal Figure 5. Test set-up for measuring dielectric constant Ta ble 6. Fresh oil prop er ties Fluid Fresh lubricating oil (SAE 15W-40) Ca pac i tance [pf] Mea sured di elec tric constant 46.36 2.03 Re sults and dis cus sion The di elec tric con stant value mea sured in the fresh SAE 15W-40 en gine lu bri cat ing oil is given in tab. 6. Fur ther, the used oil sam ples col lected at var i ous ki lo me ters per hours in ter - val from bench test en gines and fleet ve hi cles were tested, us ing the sen sor. The mea sure ment re sults of di elec tric con stant and trend anal y sis of the changes in di elec tric con stant with re spect to the changes in phys i cal prop er ties (Vis cos ity, TBN, TAN, soot) as the oil un der goes deg ra da - tion, is given in tab. 7, and plot ted in sub se quent fig ures.

S1066 Balashanmugam, V., et al.: Development of Dielectric Sensor to Monitor the Engine... Ta ble 7. Di elec tric con stant of ve hi cle oil sam ples Vehicle 1 Oil [km] 0 30000 60000 Ki ne matic vis cos ity at 100 C, cst. 14.43 12.75 12.61 TBN, mg KOH/g oil 9.64 7.6 6.24 TAN, mg KOH/g oil 2.27 3.13 3.43 Di elec tric constant 2.155 2.298 2.433 Ve hi cle 2 Oil [km] 0 30000 60000 Ki ne matic vis cos ity at 100 C, cst. 14.52 13.07 12.57 TBN, mg KOH/g oil 9.79 6.79 5.41 TAN, mg KOH/g oil 2.29 4.06 4.17 Di elec tric con stant 2.13 2.832 3.326 Ve hi cle 3 Oil [km] 0 30000 60000 Ki ne matic vis cos ity at 100 C, cst. 14.45 13.59 13.55 TBN, mg KOH/g oil 10.09 8.01 7.14 TAN, mg KOH/g oil 2.48 3.51 3.26 Di elec tric con stant 2.178 2.251 2.355 Ve hi cle 4 Oil [km] 0 30000 60000 Ki ne matic vis cos ity at 100 C, cst. 14.19 14.98 14.59 TBN, mg KOH/g oil 10.3 7.92 7.07 TAN, mg KOH/g oil 2.87 3.08 3.71 Di elec tric con stant 2.5 2.484 2.481 Figure 6. Change in TBN, TAN, viscosity, and dielectric constant with respect to oil life (oil sample from vehicle 1) The cor re la tion trend graph of di elec tric con stant with re spect to vis cos ity, to tal base num ber (TBN) and to tal acid num ber (TAN) are shown in figs. 6-9. Fig ure 6, rep re sents the di - elec tric value of 2.433 when vis cos ity, TBN and TAN are within lim its. This trend shows the re - main ing use ful life oil at 60000 km. Fig ure 7 shows the de ple tion of TBN and in crease in acid con tent, both meet ing at 5 mg KOH/g, which in - di cates the end of use ful life of oil. Ki ne matic vis cos ity and wear ma te ri als are still within lim - its. The di elec tric con stant in creases from 2.13 to 3.32 in a cor re la tion at the end of use ful life oil. Fig ure 8 rep re sents the di elec tric con stant

S1067 Figure 7. Change in TBN, TAN, viscosity and dielectric constant with respect to oil life (oil sample from vehicle 2) Figure 8. Change in TBN, TAN, viscosity and dielectric constant with respect to oil life (oil sample from vehicle 3) Figure 9. Change in TBN, TAN, viscosity and dielectric constant with respect to oil life (oil sample from vehicle 4) Figure 10. Measurement of TBN, TAN, viscosity and dielectric constant from oil sample of engine 1 value of 2.355 and the vis cos ity, TBN and TAN val ues are within limit. This shows a re main ing use ful life of oil at 60000 km. Fig ure 9 shows that the TBN, TAN, and ki ne matic vis cos ity are within lim its, and the oil has still re main ing use ful life at the end of trial. The di elec tric con stant cor re la tion re mains in the range of 2.5, while acid con tent and base num ber are within lim its. The used oil sam ples col lected from test en gines were tested. The mea sure ment re sults of di elec tric con stant and trend anal y sis of the changes in di elec tric con stant with re spect to the changes in phys i cal prop er ties (vis cos ity, TBN, TAN, soot) as the oil un der goes deg ra da tion, is given in tab. 8 and plot ted in figs. 10-13. The trend anal y sis of di elec tric con stant with re spect to vis cos ity, TBN, TAN, and soot are shown in figs. 10-13. Fig ure 13 shows that sil ica con tent has al most reached the al low able limit of 40 ppm at 450 hours, while vis cos ity, TBN, and TAN was Figure 11. Change in soot [%] in oil sample from engine 1 Figure 12. Measurement of TBN, TAN, viscosity and dielectric constant from oil sample of engine 2

S1068 Balashanmugam, V., et al.: Development of Dielectric Sensor to Monitor the Engine... within lim its. The di elec tric con stant in creases from 2.72 to 3.88 in re la tion with the end of use - ful life of oil. Figure 13. Wear debris in ol samle (silica and ferrous) from engine 2 Con clu sions An ex ten sive oil anal y sis was done based on Tribology & Lu bri cant tech nol ogy guide on SAE 15W-40 oil with die sel en gines. The 12 in - di vid ual sam ples were taken from two test en - gines and 12 in di vid ual sam ples from four on road fleet ve hi cles. These sam ples were tested for physio-chem i cal prop er ties in lab o ra tory. The anal y sis showed that the life time is vary ing for each op er a tion de pends on dif fer ent deg ra - Ta ble 8. TAN, TBN, ki ne matic vis cos ity and di elec tric con stant of used oil sam ples from test bench en gine 1 and 2 En gine 1: oil hours 0 50 100 150 200 250 Ki ne matic vis cos ity at 100 C, cst. 13.87 14.27 14.8 16 18.65 31.73 TBN, mg KOH/g 10.74 8.79 8.95 7.81 7.96 6.21 TAN mg KOH/g 3.34 3.52 4.04 3.98 3.65 4.36 Soot, [%] FT-IR 0 0.87 2.19 2.7 3.44 4.57 Di elec tric constant 2.2 2.28 2.32 2.5 2.83 3.11 En gine 2: oil hrs. 0 100 200 300 400 500 Ki ne matic vis cos ity at 100 C, cst. 13.82 13.89 13.66 13.93 13.88 13.86 TBN, mg KOH/g 8.75 8.39 7.53 8.02 8.95 8.32 TAN mg KOH/g 3.24 4.08 3.82 3.89 4.23 4.94 Di elec tric constant 2.75 2.468 2.512 2.622 3.01 3.88 da tion lev els of prop er ties such as vis cos ity, to tal base num ber, to tal acid num ber, soot and con - tam i nants. A proto type sen sor was ini tially de vel oped and val i dated in ref er ence flu ids. An er - ror of 4% was ob served. The sen sor was mod i fied with re duced gap be tween the elec trodes and with ny lon in serts. The mod i fied proto type sen sor was val i dated with ref er ence flu ids and tested in fresh en gine lu bri cat ing oil. It was used to mea sure the di elec tric con stant of used oil sam ples and cor re la tion of di elec tric con stant with re spect to changes in phys i cal/chem i cal prop er ties was stud ied. The di elec tric con stant is an ef fec tive elec tri cal prop erty to mon i tor the en gine oil deg ra da tion. A low cost sen sor of this type can mea sure the di elec tric prop erty in real time sit u a tion as well as give warn ing sig nal be fore the use ful life of oil ends. Fu ture scope of work is in this di - rec tion to have it fit ted in En gines with suit able cir cuits and warn ing de vices to re al ize the po - ten tial of pro vid ing real time in di ca tion of en gine oil life. References [1] Schmitigal, J., Moyer, S., Eval u a tion of Sen sors for On-Board Die sel Oil Con di tion Mon i tor ing of U. S. Army Ground Equip ment, US Army Jour nal, 298 (2003), 8, pp. 1-6

S1069 [2] Raadnui, S., Used Oil Deg ra da tion Detection Sensor Development, International Journal of Applied Me - chan ics and En gi neer ing, 11 (2006), 4, pp. 765-769 [3] Agoston, A., et al., Vis cos ity Sen sors for Engine Oil Condition Monitoring-Application and Interpretation of Re sults, Sen sors and Ac tu a tors A, 121 (2005), 2, pp. 327-332 [4] Zhu, J., et al., Sur vey of Lu bri ca tion Oil Condition Monitoring, Diagnostics, and Prognostics Techniques and Sys tems, Jour nal of Chem i cal Science and Technology, 2 (2013), 3, pp. 100-115 [5] Liu, Y., et al., On-Line Mon i tor ing Sys tem of Ve hi cle Lu bri cants Qual ity Based on the Permittivity, Pro - ceed ings, IEEE In ter na tional Con fer ence on Mechatronics and Au to ma tion, Changchun, China, 2009, pp. 2591-2595 [6] Tor rents, J. M., Areny, R. P., Sens ing Oil Con di tion through Tem per a ture Co ef fi cient of Di elec tric Con - stant, Pro ceed ings, XVII IMEKO World Con gress, Dubrovnik, Croatia, 2003 [7] Craft, J. W., De vel op ment of Interdigitated Elec trode Sen sors for Mon i tor ing the Di elec tric Prop er ties of Lu bri cant Oils, M. Sc. the sis, Au burn Uni ver sity, Al a bama, USA, 2010 [8] Carey, A. A., The Di elec tric Con stant of Lu bri cat ing Oils, 19898, DTIC 19980624 041 [9] Baxter, L. K., Ca pac i tive Sen sors: De sign and Ap pli ca tions, in: Series on Electronics Technology, (Ed.: R. J. Her rick), IEEE Press, New York, USA, 1997 [10] Brasseur, G., De sign Rules for Ro bust Ca pac i tive Sen sors, IEEE Transactions on Instrumentation and Mea sure ment, 52 (2003), 4, pp. 1261-1265 Paper submitted: September 13, 2015 Paper revised: January 18, 2016 Paper accepted: February 08, 2016