sensors Article
A High Performance Torque Sensor for Milling Based on a Piezoresistive MEMS Strain Gauge Yafei Qin, Yulong Zhao *, Yingxue Li, You Zhao and Peng Wang The State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, No. 28, Xianning West Road, Xi’an 710049, China;
[email protected] (Y.Q.);
[email protected] (Y.L.);
[email protected] (Y.Z.);
[email protected] (P.W.) * Correspondence:
[email protected]; Tel.: +86-29-8339-5073 Academic Editor: Vittorio M. N. Passaro Received: 29 January 2016; Accepted: 4 April 2016; Published: 9 April 2016
Abstract: In high speed and high precision machining applications, it is important to monitor the machining process in order to ensure high product quality. For this purpose, it is essential to develop a dynamometer with high sensitivity and high natural frequency which is suited to these conditions. This paper describes the design, calibration and performance of a milling torque sensor based on piezoresistive MEMS strain. A detailed design study is carried out to optimize the two mutually-contradictory indicators sensitivity and natural frequency. The developed torque sensor principally consists of a thin-walled cylinder, and a piezoresistive MEMS strain gauge bonded on the surface of the sensing element where the shear strain is maximum. The strain gauge includes eight piezoresistances and four are connected in a full Wheatstone circuit bridge, which is used to measure the applied torque force during machining procedures. Experimental static calibration results show that the sensitivity of torque sensor has been improved to 0.13 mv/Nm. A modal impact test indicates that the natural frequency of torque sensor reaches 1216 Hz, which is suitable for high speed machining processes. The dynamic test results indicate that the developed torque sensor is stable and practical for monitoring the milling process. Keywords: torque; sensitivity; frequency; MEMS strain gauge
1. Introduction In recent years, high speed machining become more and more important because of its advantages such as high production efficiency, high machining accuracy, long life of cutting tools and so on. In high speed machining processes, the cutting force directly affects the mechanical quality of the workpiece. Cutting force measurements are also an essential requirement to monitor manufacturing processes and they offer an important indicator to design machine tools, optimize machining processes, predict the surface roughness, improve the accuracy of the workpiece and detect machining tool vibrations, monitor machining cutter wear and so on. Many examples show that the cutting parameters such as cutting speed, feed rate and cutting depth often have certain deviations from the initial settings during cutting processes. It is necessary to measure the cutting force experimentally because accurate theoretical cutting force calculations cannot be performed due to the complex cutting conditions which have unsure stresses and factors. For this purpose, many dynamometers have been designed and manufactured, which are mainly based on elastic deformation of materials and used for monitoring turning and milling operations. In milling and drilling processes a table dynamometer is often typically used for monitoring the cutting force. Yaldiz et al. [1] developed a table dynamometer to measure static and dynamic cutting forces based on a strain gauge. It can measure three perpendicular cutting force components and torque. Differently, Korkut [2] also developed a three-force component dynamometer which Sensors 2016, 16, 513; doi:10.3390/s16040513
www.mdpi.com/journal/sensors
Sensors 2016, 16, 513
2 of 13
is composed of four octagonal rings placed in-house between two plates and fixed with screws. Scheer et al. [3] developed a spindle-integrated force sensor based on a piezoelectric ring for milling and drilling. Byrne et al. [4] and Park et al. [5] did a similar study with a piezoelectric force ring installed into the spindle flange, where the data could be obtained through a stator via telemetry from the rotating sensor. Totis et al. [6] also developed a rotating dynamometer to measure triaxial cutting force components in face milling. In the past, some works on spindle-integrated force sensors and rotating dynamometers based on strain gauges have been reported, and these devices could sense elemental elastic deformations with high sensitivity and output a corresponding voltage. For example, Smith et al. [7] and Suprock et al. [8] proposed a sensor integrated spindle for torque measurement; similarly, Adolfsson and Stahl [9] developed a dynamometer for measuring cutting force components at each cutting edge for face milling. Rizal et al. [10] developed an integrated rotating dynamometer based on a strain gauge which measured three cutting force components and was assembled with variety of cutting tools by the flange. There are some defects in previous research on either piezoelectric sensors or strain gauge force sensors. For piezoelectric sensors, electric charge leakage may happen during high speed milling processes and commercial piezoelectric force dynamometers (e.g., the Kistler 9257B) cost too much, even though they offer good advantages of high sensitivity, stiffness and resonant frequency. Strain gauge milling force sensors must deal with conflicting characteristics like sensitivity and natural frequency so one of them must be sacrificed in actual industrial applications. Milling torque force is recognized as a good feature for online chatter detection because of its close relationship to the material removal mechanism and its independence from the effects of tool path relative to the workpiece [11]. According to this characteristic, this paper presents a study describing an integrated torque sensor using a piezoresistive MEMS strain gauge which has small size and highly sensitivity. Polymer and silicon were usually used as materials of MEMS sensors and actuators. Liu [12] provided a comprehensive review of polymer-based MEMS, including materials, fabrication processes, and representative devices. It indicated that many polymer materials had excellent advantages for use in MEMS such as greater mechanical yield strain, significantly lower cost, being easily obtained, unique chemical, structural and biological functionalities and so on. For example, Kottapalli et al. [13,14] developed a reliable, robust, low fabrication cost and highly sensitive pressure sensor, which employed a liquid crystal polymer as a structural material. This demonstrated that polymer materials were a good choice to develop MEMS sensors for complex environments. The torque force sensing element proposed in this research is an analogue hollow cylinder structure bonded with a piezoresistive MEMS strain gauge. The MEMS strain gauge employs silicon as the substrate and is fabricated by Silicon-On-Insulator (SOI) technology, metal wires placed on the substrate are connected to a Wheatstone bridge, and the output voltage signal is transmitted by a wireless network during the milling process. This paper outlines the design and fabrication of the torque sensor, fabrication of the piezoresistive MEMS strain gauge, a method for bonding the strain gauge onto the sensing element, static calibration, model impact and dynamic tests. Experimental results indicate that the torque sensor has high performance with high sensitivity and natural frequency. The sensor developed in this work solved the principal contradiction between sensitivity and natural frequency, and the milling test results also demonstrate that the torque sensor based-MEMS strain gauge could be used to detect the stability and accuracy of the machine tool for real-time applications. 2. Design and Fabrication 2.1. Design Principle The purpose of this study was to develop a high performance sensor using piezoresistive MEMS strain gauges which can monitor milling torque forces during machining processes in real-time, as shown in Figure 1. The shear strain of the strain gauge will change when a torque force is applied to the sensing element. As a critical component, the sensing element influences the sensitivity and other properties of the whole sensor to a large extent. In this study, a thin-walled cylinder was
Sensors 2016, 16, 513
3 of 13
Sensors 2016, 16, 513
3 of 12
chosen element, as illustrated in Figure 1, and when only torque force applied to it, the strains as εx sensing and εy are tiny and we can ignore them in the calculations, thus theismain shear strain principle strains ε and ε are tiny and we can ignore them in the calculations, thus the main shear x be written y transformation can as Equation (1) under the present circumstances. According to strain transformation can beitwritten as Equation (1) under the present According to mechanical common sense, is obvious that the maximum shear straincircumstances. is produced when the angle mechanical common sense, it is obvious that the maximum shear strain is produced when the angle θ θ is ±45°, as shown in Equation (2): is ˘45˝ , as shown in Equation (2): T ε( pθq ) “xy γsin cos (1) (1) xy sin θ cos θ T
T
( 45γ)p45˝ q “ 2T 2 Gtr 2πGtr2
(2) (2)
where whereTT== torque torque force forceapplied appliedto tosensing sensingelement, element,GG== shear shear modulus modulus of of rigidity, rigidity,rr== the the outer outerradius radius of thin-walled cylinder, and t = the thickness of the sensing element. of thin-walled cylinder, and t = the thickness of the sensing element. From From Equation Equation (2), (2), itit isisobserved observed that that the the shear shear strain strain will will be begreatly greatly decreased decreased with with any anyslight slight increase of the radius under a certain torque value. The measuring sensitivity is directly proportional increase of the radius under a certain torque value. The measuring sensitivity is directly proportional to be reduced reduced with with the the increase increase of of the thedesign designparameters parametersrrand andG. G. tothe theshear shear strain strain sensitivity, sensitivity, which which will will be
Figure Figure1.1. A A diagram diagram of of the the torque torque sensor sensorusing usingaapiezoresistive piezoresistiveMEMS MEMS strain strain gauge gauge for formeasuring measuring torque force. The MEMS strain gauge is 1.8 mm ˆ 1.6 mm ˆ 0.2 mm. The strain gauge torque force. The MEMS strain gauge is 1.8 mm × 1.6 mm × 0.2 mm. The strain gauge detects detects shear shear strain with any torque force applied in a milling process. strain with any torque force applied in a milling process.
For measurement sensor, both high and high sensitivity high performance, Fora atorque torque measurement sensor, bothstiffness high stiffness and high mean sensitivity mean high so it should besoeasy to determine which designwhich parameters affect these The strain performance, it should be easy to determine designcan parameters can properties. affect these properties. sensing element can be modeled a spring appropriate stiffness.stiffness. Equation (3) shows The strain sensing element can beas modeled aselement a springwith element with appropriate Equation (3) the torsional stiffness Kms ofKthe sensing element, where G, G, t and r are defined shows the torsional stiffness ms ofstrain the strain sensing element, where t and r are definedininEquation Equation(2), (2), and andLLisisthe thelength lengthof ofthe thethin-walled thin-walledcylinder. cylinder.
2 Gtr3 2πGtr Kms K ms“ LL 3
(3) (3)
Equations Equations (2) (2) and and (3) (3) indicate indicate that that there there are are three three design design parameters parameters t,t, rr and and LL which which must must be be specified specifiedfor forthe thesensing sensingelement. element. Through Throughcomparison comparisonof ofEquations Equations(2) (2)and and(3), (3),ititisisobvious obviousthat thatthe the sensing sensingelement elementlength lengthLL should should be be minimized minimized because because itit will will increase increase the the stiffness stiffness without without affecting affecting the shear strain sensitivity. The stiffness equation shows a cube dependance on r compared the shear strain sensitivity. The stiffness equation shows a cube dependance on r compared with with its its quadratic in in the the shear strainstrain sensitivity equation. This indicates that the radius parameter quadraticdependance dependance shear sensitivity equation. This indicates that the radius rparameter should ber maximized in order toinincrease stiffness while thewhile thickness t shouldt should be as small should be maximized order to the increase the stiffness the thickness be as as possible in orderintoorder increase the sensitivity. Of course, radius the sensing small as possible to increase the sensitivity. Of neither course, the neither theofradius of theelement sensing should not be so not large interfere the milling operations nor the thickness small totoo prevent element should beassotolarge as towith interfere with the milling operations nor thetoo thickness small
to prevent buckling or damage. With comprehensive analysis of the above factors, the optimal sizes of the sensing elements are shown in Figure 2, where the outer radius, thickness, and length are 20 mm,
Sensors 2016, 16, 513
4 of 13
Sensors 2016, 16, 513
4 of 12
Sensors 2016, 513 buckling or16, damage.
4 of 12 With comprehensive analysisofofbonding the above factors, the optimal sizes the sensing 2 mm and 20 mm, respectively. For convenience with the strain gauge on theofsurface, four elements are shown in Figure 2, where the outer radius, thickness, and length are 20 mm, 2 mm and small platforms were fabricatedFor by convenience a machiningof operation. 220 mm and 20 mm, respectively. bonding with the strain gauge on the surface, four mm, respectively. For convenience of bonding with the strain gauge on the surface, four small small platforms were fabricated by a machining operation. platforms were fabricated by a machining operation.
Figure 2. Schematic view of the thin-walled cylinder. Figure 2. 2. Schematic Schematic view view of of the the thin-walled thin-walled cylinder. cylinder. Figure
In this work, 17-4PH stainless steel was chosen as sensing element material, whose elastic modulus and Poisson ratio are 2.1steel GPa andwas 0.269, respectively. Theelement surface stress of elastic the thin walled In this this work, 17-4PH stainless steel chosen as sensing material, whose elastic work, 17-4PH stainless was chosen as sensing element material, whose modulus cylinder was estimated through FEM simulation because it is difficult to obtain from a formula. The modulus andratio Poisson ratio areand 2.1 GPa 0.269, respectively. Thestress surface stress of walled the thincylinder walled and Poisson are 2.1 GPa 0.269,and respectively. The surface of the thin mesh element in this model is 2 mm. Figure 3 shows the variation of stress when the thin walled cylinder was estimated FEM simulation it is difficult to from obtaina from a formula. The was estimated through through FEM simulation becausebecause it is difficult to obtain formula. The mesh cylinder is under linear torque force. As can be seen, the outer surface stress has a good linear mesh element this model is 2Figure mm. 3Figure shows the variation stress the thincylinder walled element in thisin model is 2 mm. shows3 the variation of stressof when thewhen thin walled relationship with the applied torque force, which demonstrates that thin is suitable cylinder is under linear torque force. As can beouter seen,surface the outer surface has a relationship good linear is under linear torque force. As can be seen, the stress has walled a stress good cylinder linear as a sensing element this study. For a thin walled cylinder under torque T, the maximum relationship withtorque thefor applied torque force, which demonstrates thatthe thin walled cylinder with the applied force, which demonstrates that thin walled cylinder isforce suitable asis a suitable sensing stress change is study. distributed ±45° to thecylinder central axis, and was positive and the other as a sensing element forFor this study. Forreferring acylinder thin walled under theone torque force T, the maximum element for this aalong thin walled under the torque force T, the maximum stress change was negative. This peculiarity could be well utilized to measure stress using a Wheatstone circuit ˝ stress change along is distributed along ±45° referring theand central and oneand wasthe positive other is distributed ˘45 referring to the central to axis, one axis, was positive other and was the negative. bridge. was negative. This peculiarity could be well utilized to measure stress using a Wheatstone circuit This peculiarity could be well utilized to measure stress using a Wheatstone circuit bridge. bridge.
force. Figure 3. FEM result for the thin-walled cylinder under torque force. Figure 3. FEM result for the thin-walled cylinder under torque force.
2.2. Sensor Sensor Design Design 2.2. 2.2. Sensor Design This study study developeda high a high sensitivity natural frequency force for sensor for This developed sensitivity andand highhigh natural frequency torquetorque force sensor milling milling operations as depicted in4,Figure 4,consists which consists of three parts: 1 is a standard interface operations as depicted in Figure which of high three natural parts: Part 1 is Part a standard interface in order This study developed a high sensitivity and frequency torque force sensor for in order to fix the different types of milling cutters by a collet chuck named ER32; Part 2 is the thin to fix the different as types of milling cutters by a collet chuck named ER32; is the thin walled milling operations depicted in Figure 4, which consists of three parts: PartPart 1 is a2 standard interface walled as element sensing element with apiezoresistive bonded piezoresistive MEMS strain which cylinder used sensing a bonded MEMSnamed strain gauge which senses the in ordercylinder to fix as theused different types with of milling cutters by a collet chuck ER32; Partgauge 2 is the thin senses the stress and translates it into a voltage signal; and the third part is a standard interface stress and translates it into a voltage signal;with and the third part is a standardMEMS interface connecting with a walled cylinder used as sensing element a bonded piezoresistive strain gauge which connecting with aand HSKtranslates tool holder. HSK tool senses theholder. stress it into a voltage signal; and the third part is a standard interface connecting with a HSK tool holder.
Sensors 2016, 16, 513
5 of 13
Sensors 2016, 16, 513 Sensors 2016, 16, 513
5 of 12 5 of 12
Figure 4. Photograph of the fabricated torque sensor. Figure 4. Photograph of of the the fabricated fabricated torque torque sensor. Figure 4. Photograph sensor.
When voltage is applied to the piezoresistance to carry out dynamic monitoring in a machining When voltage is applied to the piezoresistance to carry out dynamic monitoring in a machining process, it is possible that the PN junction may produce a leakage current, and thus affect the process, itit is is possible possible thatthe thePN PNjunction junctionmay mayproduce produce a leakage current, thus affect a leakage current, andand thusSilicon-Onaffect the measurement accuracy that and performance of the sensor. In order to solve this problem, the measurement accuracy and performance of the sensor. In solve orderthis to problem, solve thisSilicon-Onproblem, measurement accuracy and performance of the sensor. In order to Insulator (SOI) was chosen as the material to fabricate our piezoresistive MEMS strain gauge. The Silicon-On-Insulator (SOI) was chosen as the material to fabricate our piezoresistive MEMS strain Insulator of (SOI) chosen as the material fabricate our piezoresistive strain gauge. The structure the was piezoresistive MEMS straintogauge adopted in this study MEMS is shown in Figure 5. It gauge. The structure of the piezoresistive MEMS strain gauge adopted inisthis study isFigure shown in structure of the piezoresistive MEMS strain gauge adopted in this study shown in 5. It consists of eight metal wire resistances, with half of them (1 KΩ each one) of the top and the Figure 5. of It consists of eight metal wire resistances, with half of(1them (1 KΩone) each of one) oftop the and top and consists eight metal wire resistances, with half of them KΩ each the the remaining ones underneath (333 Ω each one). Taking the top as an example, four resistances can be the remaining ones underneath (333 Ω each one). Takingthe thetop topasasan an example,four four resistances can be remaining ones underneath (333 Ω each one). Taking used to organize an independent Wheatstone circuit bridge, where example, resistances R1 resistances and R4 along the used to organize an independent Wheatstone circuit bridge, where resistances R1 and R4 along the the [110] direction and R2 and R3 in the [ 1 10 ] direction are included. Due to the excellent mechanical the [110] direction and R2 and R3 in in the the [[110] are to the excellent mechanical ] direction direction are included. included. Due Due 1 10 plane and electrical properties of the [100] crystal of monocrystalline silicon, a piezoresistive strain and electrical properties of the [100] crystal plane of monocrystalline silicon, a piezoresistive plane monocrystalline piezoresistive strain gauge whose size is 1.8 mm × 1.6 mm crystal × 0.2 mm has of been fabricated by MEMS technology. gauge whose size is 1.8 mm ˆ 1.6 mm mm ׈0.2 0.2mm mmhas hasbeen beenfabricated fabricatedby byMEMS MEMStechnology. technology. × 1.6
Figure 5. Schematic view of the designed MEMS strain gauge. Figure 5. 5. Schematic of the the designed designed MEMS MEMS strain strain gauge. gauge. Figure Schematic view view of
The piezoresistive MEMS strain gauge was processed mainly through the following steps as The piezoresistive MEMS strain gauge was processed mainly through the following steps as shown inpiezoresistive Figure 6. First,MEMS a Silicon-On-Insulator is cleaned with hydrogen fluoride solution; The strain gauge waswafer processed mainly through the following steps as shown in Figure 6. First, a Silicon-On-Insulator wafer is cleaned with hydrogen fluoride solution; second, weFigure performed p-type boron ion doping on the SOI wafer surface with an fluoride ion implantation shown in 6. First, a Silicon-On-Insulator wafer is cleaned with hydrogen solution; second, we performed p-type boron ion doping on the SOI wafer surface with an ion implantation system. that, an annealing treatment in a 1100 °C nitrogen environment maintained for one second, After we performed p-type boron ion doping on the SOI wafer surface withisan ion implantation system. After that, the an annealing treatment in a 1100 ˝°C nitrogen environment is maintained for one hour to distribute mixed concentration evenly environment on the surface; then, we carry out system. After that, an annealing treatment inofaboron 1100 ion C nitrogen is maintained for one hour to distribute the mixed concentration of boron ion evenly on the surface; then, we carry out positive photolithography and dense ion implantation of boron for the ohmic contact area in order hour to distribute the mixed concentration of boron ion evenly on the surface; then, we carry out positive photolithography and dense ion implantation of boron for the ohmic contact area in order to form aphotolithography low resistance ohmic area; in the fourth areas are to positive and contact dense ion implantation of step, boronthe forremaining the ohmicSi contact areaetched in order to form low resistance ohmic contact area; in theohmic fourth step, the remaining Si areas are etched to form theaaembossed resistance and graphics contact arearemaining on both sides of the to form low resistance ohmic contact area;ofinthe the fourth step, the Si areas areresistance; etched to form the embossed resistance andby graphics the ohmic contactprocess; area on both sides of the fifthly, weembossed obtain SiO 2 thin films using aof oxidation using lowresistance; pressure form the resistance and graphics ofthermal the ohmic contact area on sixthly, both sides of the resistance; fifthly, we obtain SiO 2 thin films by using a thermal oxidation process; sixthly, using low pressure chemical vapor deposition (LPCVD) Si3N4 thin films are formed; seventh, positive photolithography chemical vapor (LPCVD) 3N4 thin films are formed; seventh, positive photolithography is performed so deposition as to form wire holesSi between the aluminium wire and resistance; eighth, we obtain is performed so as to form wire holes between the aluminium wire and resistance; eighth, we obtain
Sensors 2016, 16, 513
6 of 13
fifthly,2016, we 16, obtain Sensors 513 Sensors 2016, 16, 513
SiO2 thin films by using a thermal oxidation process; sixthly, using low pressure 6 of 12 6 of 12 chemical vapor deposition (LPCVD) Si3 N4 thin films are formed; seventh, positive photolithography performed to form vapor wire holes between aluminium wire and resistance; eighth, ais Ti/Al layer so byasphysical deposition onthe the positive silicon wafer and then formwe theobtain metala a Ti/Al layer by physical vapor deposition on the positive silicon wafer and then form the metal Ti/Al layerleading by physical deposition on the positive silicon wafer and then formusing the metal electrode wirevapor tunnel by Inductively Coupled Plasma (ICP); finally, ICP electrode etching, electrode leading wire tunnel by Inductively Coupled Plasma (ICP); finally, using ICP etching, leadingmetal wire tunnel by Inductively Plasma (ICP); finally, ICPstrain etching, useless useless is removed. Figure 7a Coupled shows a SEM photograph of theusing MEMS gauge, and metal some useless metal is removed. Figure 7a shows a SEM photograph of the MEMS strain gauge, and some is removed. Figure 7a shows SEM photograph the MEMS strain gauge, and some as-fabricated as-fabricated strain gauges areapresented in Figureof7b. as-fabricated strain gauges are presented in Figure 7b. strain gauges are presented in Figure 7b.
Figure 6. 6. Fabrication Fabrication process process of of MEMS MEMS strain strain gauge. gauge. Figure Figure 6. Fabrication process of MEMS strain gauge.
(a) (a)
(b) (b)
Figure 7. (a) SEM of MEMS strain gauge; (b) aa photograph of MEMS strain gauge. Figure Figure7.7. (a) (a) SEM SEM of ofMEMS MEMSstrain straingauge; gauge;(b) (b) aphotograph photographof ofMEMS MEMSstrain straingauge. gauge.
2.3. Bonding 2.3. 2.3. Bonding Bonding Bonding isanother another critical aspect strongly influences outputofsensitivity of Bonding critical aspect whichwhich strongly influences the outputthe sensitivity Bondingis is another critical aspect which strongly influences the output piezoresistive sensitivity of piezoresistive MEMS Several strain gauges. Several successful bonding methods based on silicon on the MEMS strain gauges. bonding methods based on silicon on the surface of stainless piezoresistive MEMS strain successful gauges. Several successful bonding methods based on silicon on the surface of stainless steel substrates were proposed [15–19]. steel substrates weresteel proposed [15–19]. surface of stainless substrates were proposed [15–19]. The piezoresistive MEMS strain gauge may rotate away from the original direction or tilt [20] The The piezoresistive piezoresistive MEMS MEMS strain strain gauge gauge may may rotate rotate away away from from the the original original direction direction or or tilt tilt [20] [20] which could cause strain measurement errors. Bonding directions could be controlled reasonably by which measurement errors. errors.Bonding Bondingdirections directionscould could controlled reasonably which could could cause strain measurement bebe controlled reasonably by marking out the positioning line on the surface of sensing element. In addition, highly technical by marking out the positioning line on the surface of sensing element. In addition, highly technical marking out the positioning line on the surface of sensing element. addition, highly technical requirements in manufacturing sensing elements such as flatness, roughness could ensure the steel requirements requirements in inmanufacturing manufacturing sensing sensing elements elements such such as as flatness, flatness, roughness roughness could could ensure ensure the thesteel steel surface is clean and flat. In this case, the strain measurement sensitivity errors caused by the tilt are surface surfaceisisclean cleanand andflat. flat. In In this thiscase, case,the thestrain strainmeasurement measurement sensitivity sensitivity errors errors caused caused by by the the tilt tilt are are expected to be reduced to aa minimal impact. expected to be reduced to minimal impact. expected to be reduced to a minimal impact. The piezoresistive MEMS gauge and PCB with a circular through-hole in theincentre were The MEMSstrain strain gauge and PCB with a circular through-hole the centre The piezoresistive MEMS strain gauge and PCB with a circular through-hole in the centre were bonded to the stainless steel surface using M-Bond 610 epoxy, which is a solvent-thinned, epoxywere bonded the stainless steel surface M-Bond 610 epoxy, is a solvent-thinned, bonded to thetostainless steel surface using using M-Bond 610 epoxy, which which is a solvent-thinned, epoxyphenolic adhesive for high-performance applications. As aAsfirst step, thethegauging was epoxy-phenolic adhesive for high-performance applications. a first step, gaugingarea phenolic adhesive for high-performance applications. As a first step, the gauging area was was thoroughly degreased with GC-6 isopropyl alcohol, and a ballpoint pen was used to draw alignment thoroughly degreased with GC-6 isopropyl alcohol, and a ballpoint pen was used to draw alignment marks on the sensing element surface in order to make sure that strain gauge is installed in the right marks on the sensing element surface in order to make sure that strain gauge is installed in the right place, then it was cleaned with ethyl alcohol and wiped dry with a gauze sponge. Secondly, M-Bond place, then it was cleaned with ethyl alcohol and wiped dry with a gauze sponge. Secondly, M-Bond 610 was applied to the sensing element surface bonding area and the back of the strain gauge and 610 was applied to the sensing element surface bonding area and the back of the strain gauge and
Sensors 2016, 16, 513
7 of 13
thoroughly degreased with GC-6 isopropyl alcohol, and a ballpoint pen was used to draw alignment marks on the sensing element surface in order to make sure that strain gauge is installed in the Sensorsplace, 2016, 16, 513 it was cleaned with ethyl alcohol and wiped dry with a gauze sponge. Secondly, 7 of 12 right then M-Bond 610 was applied to the sensing element surface bonding area and the back of the strain gauge PCB with a circular through-hole in centre, and the assembly was set each aside to air-dry for at least and PCB with a circular through-hole in centre, and the assembly was set each aside to air-dry for 15 min. Then the piezoresistive MEMS strain gauge and PCB assembly was returned to its original at least 15 min. Then the piezoresistive MEMS strain gauge and PCB assembly was returned to its position over the layout marks. The assembly of strain gauge and PCB was tacked down with a original position over the layout marks. The assembly of strain gauge and PCB was tacked down with certain amount of pressure. Spring clamps were used to apply a pressure of about 200 kN/m2 during a certain amount of pressure. Spring clamps were used to apply a pressure of about 200 kN/m2 during the curing cycle. Then the clamped sensor was place into an oven and the temperature raised to 125 °C the curing cycle. Then the clamped sensor was place into an oven and the temperature raised to 125 ˝ C at a rate of approximately 10 °C per minute and held there for 2 h. Step 2 must be completed within at a rate of approximately 10 ˝ C per minute and held there for 2 h. Step 2 must be completed within 4 h. After remaining for 2 h in 125 °C, the sensor should be naturally cooled to room temperature. 4 h. After remaining for 2 h in 125 ˝ C, the sensor should be naturally cooled to room temperature. Finally, the sensor is placed in the oven and the temperature raised to 160 °C and maintained 2 h and Finally, the sensor is placed in the oven and the temperature raised to 160 ˝ C and maintained 2 h then cooled to room temperature in a natural environment repeatedly. The bonding and curing and then cooled to room temperature in a natural environment repeatedly. The bonding and curing treatment are completed by these steps. treatment are completed by these steps.
3. Static 3. Static Calibration Calibration In this this work, work, the the testing testing apparatus apparatus used used to to measure measure the the stress stress are are shown shown in inFigure Figure8a. 8a. They They In include a CNC testing machine (WNZ-200, XI'AN LETRY, Xi’an, China) that could provide constant include a CNC testing machine (WNZ-200, XI’AN LETRY, Xi’an, China) that could provide constant torque force, force, aa digital digital multimeter multimeter (8846A, (8846A, FLUKE FLUKE CORPORATION, CORPORATION, Everett, Everett, WA, WA,USA) USA)which which could could torque gauge voltage voltage and and aa power power supply supply (GPS-3303C, (GPS-3303C, GWINSTEK GWINSTEK Electronic Electronic Technology Technology Co. Co. Ltd, Ltd, Suzhou, Suzhou, gauge China) which could apply an excitation voltage of 5V DC for the torque sensor. To operate the testing China) which could apply an excitation voltage of 5V DC for the torque sensor. To operate the apparatus, a torque force was applied to the end face by the electronic force regulator. The testing apparatus, a torque force was applied to the end face by the electronic force regulator. The measurement range range of of the the torque torque force force isis expected expected to to be be0–40 0–40Nm. Nm. The The applied applied torque torque force force was was measurement raised from zero to its maximum value in steps of 4 Nm, and holding time was maintained for 20 to raised from zero to its maximum value in steps of 4 Nm, and holding time was maintained for 20 to 40 ss in ineach eachstep. step. The The measuring measuring circuits circuits were were excited excited by by the the power power supply supply and and the the output output voltage voltage 40 were recorded by the high-accuracy and high-resolution digital multimeter. Five cycles including were recorded by the high-accuracy and high-resolution digital multimeter. Five cycles including loading and and unloading unloading procedures procedures were were implemented, implemented, and and calibration calibration result result of of the the torque torque force force is is loading depicted in in Figure Figure 8b. 8b. As As shown shown in in the the figure, figure, the the calibration calibration data data have have been been described described with with aa red red depicted solid line obtained by linear fitting and error bars. Three calibration values with error bars were taken solid line obtained by linear fitting and error bars. Three calibration values with error bars were as examples, and respectively partially magnified. The experimental sensitivity of the of torque sensor taken as examples, and respectively partially magnified. The experimental sensitivity the torque voltagevoltage signal signal (un-amplified) is 0.13 mv/Nm. The The developed sensor static sensor (un-amplified) is 0.13 mv/Nm. developed sensorpossesses possesses favorable favorable static properties with a linearity error of less than 1.6%, which implies that it can meet the goal of high properties with a linearity error of less than 1.6%, which implies that it can meet the goal of high accuracymeasurement. measurement. accuracy
Figure 8. (a) Experimental setup for static calibration; (b) Calibration curve for the torque force. Figure 8. (a) Experimental setup for static calibration; (b) Calibration curve for the torque force.
4. Model Impact Test The system response of the sensor reflects its dynamic performance, which needs to be considered when the torque sensor is installed on the spindle monitoring machine tool in the cutting process. The natural frequency of the sensor affects the dynamic response of the system, which must be much higher than the frequency produced by the cutting tool in order to ensure that the collected signals would not be disturbed during the cutting process. Generally, the natural frequency should
Sensors 2016, 16, 513
8 of 13
4. Model Impact Test The system response of the sensor reflects its dynamic performance, which needs to be considered when the torque sensor is installed on the spindle monitoring machine tool in the cutting process. The natural frequency of the sensor affects the dynamic response of the system, which must be much Sensors2016, 2016, 16, 513frequency produced by the cutting tool in order to ensure that the collected signals of12 12 higher than the Sensors 16, 513 88of would not be disturbed during the cutting process. Generally, the natural frequency should be at least be at attimes least larger four times times larger than the thecaused frequency caused bytool. the cutting cutting tool.the Tonatural identify the natural natural be least four than frequency caused by the tool. To identify the four than larger the frequency by the cutting To identify frequency of frequency of the torque sensor under actual working conditions, the natural frequency of the torque frequency the torque under actual working the natural frequency of the system torque the torque of sensor under sensor actual working conditions, theconditions, natural frequency of the torque sensor sensor systemby was obtained by aa modal modal impcting test. The torque sensor was excited by hammer modal sensor system was obtained by impcting The torque sensor excited by aa modal was obtained a modal impcting test. The torquetest. sensor was excited by awas modal impact impact hammer (type 086D05), and aa triaxial triaxial vibration transducer (type 95663) was connected to the the impact hammer (type 086D05), and vibration 95663) to (type 086D05), and a triaxial vibration transducer (type transducer 95663) was(type connected towas the connected sensing element, sensing element, as illustrated in Figure 9. The signals excited by hammer and vibration transducer sensing element, as illustrated in Figure 9. The excited by hammer and vibration transducer as illustrated in Figure 9. The signals excited bysignals hammer and vibration transducer were acquired by were acquired by a data acquisition and modal analysis system (type SCADAS305) manufactured by were acquired by a data acquisition and modal analysis system (type SCADAS305) manufactured by a data acquisition and modal analysis system (type SCADAS305) manufactured by LMS Company LMS Company (Leuven, Belgium). Finally, the natural frequencies of the sensor in three directions LMS Company (Leuven, Finally, the natural of the sensor in three directions (Leuven, Belgium). Finally,Belgium). the natural frequencies of thefrequencies sensor in three directions were calculated by were calculated by the LMS LMSFigure Test Lab Lab software. Figure 10 shows shows the the amplitude-frequency amplitude-frequency functions were calculated by the Test 10 functions the LMS Test Lab software. 10 software. shows theFigure amplitude-frequency functions in three directions. inisthree three directions. is obvious obvious that the theof natural frequencies of the the sensor sensor are1248 approximately 1216, in directions. is that natural frequencies of are approximately 1216, It obvious that theItItnatural frequencies the sensor are approximately 1216, and 2362 Hz. In 1248 and 2362 Hz. In this study, cutting tests were performed using a two toothed end mill. On 1248study, and 2362 Hz. tests In this study, cutting using tests were using a two toothed of end On this cutting were performed a twoperformed toothed end mill. On account themill. lowest accountfrequency of the the lowest lowest natural frequency of 1216 Hz, this sensor cancutting be used used for dynamic dynamic cuttingbelow force account of natural 1216 this can be for cutting force natural of 1216 Hz,frequency this sensorof can beHz, used forsensor dynamic force measurements measurements below 1126/4 × ½½a four 60 toothed 8445 rpm. rpm. four toothed toothed end mill were were chosen as cutting cutting measurements below 1126/4 ×× 60 == 8445 IfIf aa four end mill as 1126/4 ˆ ½ ˆ 60 = 8445 rpm. ×If end mill were chosen as cutting tool, chosen the spindle speed tool, the spindle speed would be limited to 4220 rpm in the actual machining process. tool, the would bethe limited tomachining 4220 rpm in the actual machining process. would bespindle limitedspeed to 4220 rpm in actual process.
Figure 9. 9. Experimental Experimental setup setup of of the the modal modal impact tests. Figure impact tests.
Figure 10. 10. Frequency Frequency and and amplitude amplitude result result of of the theimpacting impactingmodal modaltest. test. Figure 10. Frequency and amplitude result Figure of the impacting modal test.
5. Cutting Cutting Test Test and and Discussion Discussion 5. Dynamic cutting cutting tests tests were were performed performed in in order order to to evaluate evaluate the the performance performance of of the the developed developed Dynamic sensor in in real real operation. operation. The The cutting cutting tests tests were were operated operated on on aa numerical numerical control control milling milling machine machine sensor under dry cutting conditions as shown in Figure 11. The cutting tool was a two toothed 16 mm mm under dry cutting conditions as shown in Figure 11. The cutting tool was a two toothed 16 diameter end mill and the workpiece material was 7075 aluminum. These experiments were diameter end mill and the workpiece material was 7075 aluminum. These experiments were performedwith withaaradial radialcutting cuttingdepth depthof of100 100percent percentimmersion, immersion,axial axialcutting cuttingdepth depthof of11mm, mm,feedrate feedrate performed
Sensors 2016, 16, 513
9 of 12
receiver was also used. The acquired milling torque force signals areare given inin Figures 12–14. The top receiver was also used. The acquired milling torque force signals given Figures 12–14. The top graph shows the dynamic milling torque forces inin the time domain. Their wave form inin the frequncy graph shows the dynamic milling torque forces the time domain. Their wave form the frequncy domain is is depicted inin the bottom graph. AllAll the spectra areare respectively normalized byby their own domain depicted the bottom graph. the spectra respectively normalized their own Sensors 2016, 16, 513 9 of 13 maximum values. maximum values. Figure 1212 shows the results ofof measured milling torque forces with a spindle speed ofof 500 rpm, Figure shows the results measured milling torque forces with a spindle speed 500 rpm, where the measured frequecy generated by the milling tool tooth is 17.2 Hz. Figures 13 and 14 show where the measured frequecy generated by the milling tool tooth is 17.2 Hz. Figures 13 and 14 show 5. Cutting Test and Discussion the theresults resultsthat thatwhen whenthe thespindle spindlespeeds speedsareareincreased increasedtoto1000 1000rpm rpmand and2000 2000rpm, rpm,whereby wherebythe the Dynamic cutting tests were performed in order to evaluate the performance of the developed detected frequency also increased to 32.9 Hz and 69.6 Hz. During the milling process, because the detected frequency also increased to 32.9 Hz and 69.6 Hz. During the milling process, because the sensor intool real operation. Theteeth, cutting tests were operated on derived a numerical control milling machine under cutting has two cutting the measured frequency from torque sensor approximately cutting tool has two cutting teeth, the measured frequency derived from torque sensor approximately dry cutting conditions as speed. shown inAll Figure 11. The cutting tool was a the two toothed of 16of mm diameter equals twice the spindle All the spectra areare normalized byby the amplitude the maximum equals twice the spindle speed. the spectra normalized amplitude the maximum end mill and the workpiece material was 7075 aluminum. These experiments were performed with a frequency peak seen in the respective spectrum. frequency peak seen in the respective spectrum. radialAcutting depth of 100 percent immersion, axial cutting depth of 1 mm, feedrate of 0.15 mm per milling test record with the same milling parameters as mentioned above is presented inin A milling test record with the same milling parameters as mentioned above is presented tooth under steady state milling conditions with allwas teeth in place. The spindle speed was separately Figure 15,15, except that one ofof the milling tool teeth broken and the rotation speed was 800 rpm. Figure except that one the milling tool teeth was broken and the rotation speed was 800 rpm. set at 500, 1000 and 2000 rpm. During the experiments, the sampling rate of the cutting torque forces The irregular signal obviously indicates the effect of the broken milling tool tooth in the time domain. The irregular signal obviously indicates the effect of the broken milling tool tooth in the time domain. measurement was 10 kHz, andsituation a situation pair of wireless signal processing andthrough transmitter and receiver waswhen also It Itmeans the abnormal will found immediately time domain meansthat that the abnormal willbebe found immediately throughthethe time domain when used. The acquired milling torque force signals are given in Figures 12–14. The top graph shows thethe monitoring the machining process. Certainly, these frequencies have to be lower than a quarter of monitoring the machining process. Certainly, these frequencies have to be lower than a quarter of dynamic milling torque forces in the time domain. Their wave form in the frequncy domain is depicted natural frequency (1216 Hz). By comparing the measured milling torque signals, when the spindle natural frequency (1216 Hz). By comparing the measured milling torque signals, when the spindle in the bottom graph. All the spectra are respectively normalized by their ownvalue maximum values. speed increases with other milling parameters invariant, the measured torque decreases slowly. speed increases with other milling parameters invariant, the measured torque value decreases slowly.
Figure 11.11. Dynamic cutting experiment. Figure Dynamic cutting experiment. Figure 11. Dynamic cutting experiment.
Figure 12. Plots of of torque force signals in time and frequency domain at a spindle speed of of 500 rpm. Figure Plots torque force signals time and frequency domain a spindle speed 500 rpm. Figure 12.12. Plots of torque force signals in in time and frequency domain at aatspindle speed of 500 rpm.
Sensors 2016, 16, 513 Sensors Sensors2016, 2016,16, 16,513 513
10 of 12 1010ofof1312
Figure 13. 13. Plots of of torque force signals spindle speed speedof of1000 1000rpm. rpm. Figure Plots torque force signalsinintime timeand andfrequency frequencydomain domain at a spindle Figure 13. Plots of torque force signals in time and frequency domain at a spindle speed of 1000 rpm.
Figure 14. Plots of torque force signals in time and frequency domain at a spindle speed of 2000 rpm. Figure 14. Plots of torque force signals in time and frequency domain at a spindle speed of 2000 rpm. Figure 14. Plots of torque force signals in time and frequency domain at a spindle speed of 2000 rpm.
Figure 12 shows the results of measured milling torque forces with a spindle speed of 500 rpm, where the measured frequecy generated by the milling tool tooth is 17.2 Hz. Figures 13 and 14 show the results that when the spindle speeds are increased to 1000 rpm and 2000 rpm, whereby the detected frequency also increased to 32.9 Hz and 69.6 Hz. During the milling process, because the cutting tool has two cutting teeth, the measured frequency derived from torque sensor approximately equals twice the spindle speed. All the spectra are normalized by the amplitude of the maximum frequency peak seen in the respective spectrum. A milling test record with the same milling parameters as mentioned above is presented in Figure 15, except that one of the milling tool teeth was broken and the rotation speed was 800 rpm.
Sensors 2016, 16, 513
11 of 13
The irregular signal obviously indicates the effect of the broken milling tool tooth in the time domain. It means that the abnormal situation will be found immediately through the time domain when monitoring the machining process. Certainly, these frequencies have to be lower than a quarter of the natural frequency (1216 Hz). By comparing the measured milling torque signals, when the spindle Sensorsincreases 2016, 16, 513 11 of 12 speed with other milling parameters invariant, the measured torque value decreases slowly.
Figure 15. Plots of torque force signals in time and frequency domain at a spindle speed of 800 rpm. Figure 15. Plots of torque force signals in time and frequency domain at a spindle speed of 800 rpm.
6.6. Conclusions Conclusions In Inthis thisstudy, study,an aninnovative innovativepiezoresistive piezoresistiveMEMS MEMSstrain straingauge-based gauge-basedtorque torquesensor sensorwith withboth both high highsensitivity sensitivityand andhigh highnatural naturalfrequency frequencywas wasdeveloped developedfor formilling millingprocesses. processes.AApiezoresistive piezoresistive MEMS MEMSstain staingauge gaugewas wasemployed employedto tomeasure measuretorque torqueininmilling millingprocesses processesfor forthe thefirst firsttime. time.Cutting Cutting tools are interchangeable so the developed sensor can support a variety of machining processes. tools are interchangeable so the developed sensor can support a variety of machining processes.The The static staticexperiments experimentsshowed showedthat thatthe thesensitivity sensitivitywas wasapproximately approximately0.13 0.13mv/Nm, mv/Nm,which whichisismuch muchhigher higher than torque rotating dynamometer. TheThe modal impacting test test results showshow that the thanthat thatofofa atraditional traditional torque rotating dynamometer. modal impacting results that natural frequency of the sensor reaches 1216 Hz, which means that the dynamic operation range the the natural frequency of the sensor reaches 1216 Hz, which means that the dynamic operationof range torque sensor sensor is suitable for spindle speedsspeeds of lessofthan machining processes whenwhen the of the torque is suitable for spindle less8445 than rpm 8445in rpm in machining processes cutting tool has two teeth. It is observed that the piezoresistive MEMS strain gauge-based torque sensor the cutting tool has two teeth. It is observed that the piezoresistive MEMS strain gauge-based torque primely picks up picks the periodic the cutting in the experiments. experimental sensor primely up thefrequency periodic of frequency of tooth the cutting tooth in the The experiments. The results indicate that the torque sensor based on a MEMS chip developed in this work optimizes the experimental results indicate that the torque sensor based on a MEMS chip developed in this work intrinsic contradiction between senstivity and natural frequency in traditional strain sensors. It optimizes the intrinsic contradiction between senstivity and natural frequency in gauge traditional strain isgauge feasible to use It torque sensortomonitoring the stability of milling sensors. is feasible use torqueof sensor monitoring of theprocesses stability in of real-time. milling processes in It is worth noting that the measured piezoresistive MEMS sensor signal is a slightly fluctuant real-time. version of the milling torque, especially the signal wave was anomalous when the milling tool was It is worth noting that the measured piezoresistive MEMS sensor signal is a slightly fluctuant damaged asthe shown in Figure 15,especially so this variation can wave be used foranomalous monitoringwhen the stability of machining version of milling torque, the signal was the milling tool was processes under high speed cutting conditions. The variations of the measured signal could be used toof damaged as shown in Figure 15, so this variation can be used for monitoring the stability reveal wear or breakageunder of thehigh cutting teeth, to detect the occurrence of chatter,oftothe monitor changes in machining processes speed cutting conditions. The variations measured signal cutting parameters such as feed rate, cutting depth, spindle rotation speed and so on. In conclusion, could be used to reveal wear or breakage of the cutting teeth, to detect the occurrence of chatter, itto would significantly processes monitoring and depth, controlling applications. monitor changes inbenefit cuttingmachining parameters such as in feed rate, cutting spindle rotation speed and
so on. In conclusion, it would significantly benefit machining processes in monitoring and controlling
Acknowledgments: This research is supported by the National High Technology Research and Development applications. Program of China (Grant No. 2013AA041108), National Natural Science Foundation of China (Grant No. 51421004), The National Science Fund for Distinguished Young Scholars (Grant No. 51325503), Changjiang Scholars and Acknowledgments: This research is supported by the National High Technology Research and Development Program of China (Grant No. 2013AA041108), National Natural Science Foundation of China (Grant No. 51421004), The National Science Fund for Distinguished Young Scholars (Grant No. 51325503), Changjiang Scholars and Innovative Research Team in University of China (No. IRT_14R45), The Funds for Creative Research Groups of China (No. 51421004).
Author Contributions: The authors appreciate the help of Yulong Zhao from The State Key Laboratory for
Sensors 2016, 16, 513
12 of 13
Innovative Research Team in University of China (No. IRT_14R45), The Funds for Creative Research Groups of China (No. 51421004). Author Contributions: The authors appreciate the help of Yulong Zhao from The State Key Laboratory for Manufacturing Systems Engineering for his support and guidance in his research; Yingxue Li helped collected the milling data and contribute to draft amendment; You Zhao helped with the experimental preparation; Peng Wang helped implement the dynamic testing; Yafei Qin designed the experimental program, took charge in testing progress, performed data analysis and wrote the manuscript. Conflicts of Interest: The authors declare that they have no conflict of interest.
References 1.
2. 3.
4. 5. 6. 7. 8. 9. 10. 11. 12. 13.
14.
15. 16.
17. 18.
Yaldız, S.; Unsacar, F.; Saglam, H.; Isık, H. Design, development and testing of a four-component milling dynamometer for the measurement of cutting force and torque. Mech. Syst. Signal Process. 2007, 21, 1499–1511. [CrossRef] Korkut, I. A dynamometer design and its construction for milling operation. Mater. Des. 2003, 24, 631–637. [CrossRef] Hoffmann, P.; Scheer, C.; Zurich, E. Spindle-integrated force sensors for monitoring drilling and milling processes. In Proceedings of the 9th International Conference for Sensors, Transducers & Systems, Nurnberg, Germany, 18–20 May 1999; pp. 1–7. Byrne, G.; O’Donnell, G.E. An integrated force sensor solution for process monitoring of drilling operations. CIRP Ann.-Manuf. Technol. 2007, 56, 89–92. [CrossRef] Park, S.S. High Frequency Bandwidth Cutting Force Measurements in Milling Using the Spindle Force Sensor System. Ph.D. Thesis, University of British Columbia, Vancouver, BC, Canada, 2004. Totis, G.; Wirtz, G.; Sortino, M.; Klocke, F. Development of a dynamometer for measuring individual cutting edge forces in face milling. Mech. Syst. Signal Process. 2010, 24, 1844–1857. [CrossRef] Smith, D.A.; Smith, S.; Tlusty, J. High performance milling torque sensor. J. Manuf. Sci. Eng. 1998, 120, 504–514. [CrossRef] Suprock, C.A.; Nichols, J.S. A low cost wireless high bandwidth transmitter for sensor-integrated metal cutting tools and process monitoring. Int. J. Mechatron. Manuf. Syst. 2009, 2, 441–454. [CrossRef] Adolfsson, C.; Ståhl, J.E. Cutting force model for multi-toothed cutting processes and force measuring equipment for face milling. Int. J. Mach. Tools Manuf. 1995, 35, 1715–1728. [CrossRef] Rizal, M.; Ghani, J.A.; Nuawi, M.Z.; Che Haron, C.H. Development and testing of an integrated rotating dynamometer on tool holder for milling process. Mech. Syst. Signal Process. 2015, 52, 559–576. [CrossRef] Kuljanic, E.; Sortino, M.; Totis, G. Multisensor approaches for chatter detection in milling. J. Sound Vib. 2008, 312, 672–693. [CrossRef] Liu, C. Recent developments in polymer MEMS. Adv. Mater. 2007, 19, 3783–3790. [CrossRef] Kottapalli, A.G.P.; Asadnia, M.; Miao, J.M.; Barbastathis, G.; Triantafyllou, M.S. A flexible liquid crystal polymer MEMS pressure sensor array for fish-like underwater sensing. Smart Mater. Struct. 2012, 21, 115030. [CrossRef] Kottapalli, A.G.P.; Tan, C.W.; Olfatnia, M.; Mao, J.M.; Barbastathis, G.; Triantafyllou, M.S. A liquid crystal polymer membrane MEMS sensor for flow rate and flow direction sensing applications. J. Micromech. Microeng. 2011, 21, 085006. [CrossRef] Ferri, M.; Mancarell, F.; Belsito, L.; Roncaglia, A.; Yan, J.; Seshia, A.A.; Soga, K.; Zalesky, J. Strain sensing on steel surfaces using vacuum packaged MEMS resonators. Procedia Eng. 2010, 5, 1426–1429. [CrossRef] Ferri, M.; Belsito, L.; Mancarella, F.; Masini, L.; Roncaglia, A.; Yan, J.; Seshia, A.A.; Zalesky, J.; Soga, K. Fabrication and testing of a high resolution extensometer based on resonant MEMS strain sensors. In Proceedings of the 16th International Solid-State Sensors, Actuators and Microsystems Conference (Transducers), Beijing, China, 5–9 June 2011; pp. 1056–1059. Sosnowchik, B.; Azevedo, R.; Myers, D.; Chan, M.; Pisano, A.; Lin, L. Rapid siliconto-steel bonding by induction heating for MEMS strain sensors. J. Microelectro Mech. Syst. 2012, 21, 497–506. [CrossRef] Sosnowchik, B.D.; Azevedo, R.G.; Cao, A.; Lin, L.; Pisano, A.P. Silicon-to-steel bonding using rapid thermal annealing. IEEE Trans. Adv. Packag. 2005, 28, 626–634. [CrossRef]
Sensors 2016, 16, 513
19.
20.
13 of 13
Kim, Y.D.; Lee, C.S.; Kwon, S.J. Novel 3D force sensor using ultra-thin silicon strain gauge bonded on metal membrane. In Proceedings of the 15th International Solid-State Sensors, Actuators and Microsystems Conference (Transducers), Denver, CO, USA, 21–25 June 2009; pp. 920–1923. Myers, D.R.; Chan, M.W.; Vigevani, G.; Sosnowchik, B.D.; Azevedo, R.G.; Jog, A.V.; Lin, L.; Pisano, A.P. Torque measurements of an automotive halfshaft using micro double-ended tuning fork strain gauges. Sens. Actuators A Phys. 2013, 204, 79–87. [CrossRef] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
