Professional Documents
Culture Documents
Full download An experimental investigation of the effects of the compliant joint method on feedback compensation of pre-sliding/pre-rolling friction Xin Dong & Chinedum E. Okwudire file pdf all chapter on 2024
Full download An experimental investigation of the effects of the compliant joint method on feedback compensation of pre-sliding/pre-rolling friction Xin Dong & Chinedum E. Okwudire file pdf all chapter on 2024
Full download An experimental investigation of the effects of the compliant joint method on feedback compensation of pre-sliding/pre-rolling friction Xin Dong & Chinedum E. Okwudire file pdf all chapter on 2024
https://ebookmass.com/product/classification-of-mesoscopic-
tribological-properties-under-dry-sliding-friction-for-
microforming-operation-tetsuhide-shimizu-ming-yang-ken-ichi-
manabe/
https://ebookmass.com/product/effects-of-sand-compaction-on-
liquefaction-during-the-tokachioki-earthquake-yorihiko-ohsaki/
The Testament of Lazarus: The Pre-Christian Gospel of
John Janet Tyson
https://ebookmass.com/product/the-testament-of-lazarus-the-pre-
christian-gospel-of-john-janet-tyson/
https://ebookmass.com/product/offshore-compliant-platforms-
analysis-design-and-experimental-studies-1-edition-srinivasan-
chandrasekaran/
https://ebookmass.com/product/method-substance-and-the-future-of-
african-philosophy-1st-edition-edwin-e-etieyibo-eds/
https://ebookmass.com/product/wittgenstein-on-logic-as-the-
method-of-philosophy-re-examining-the-roots-and-development-of-
analytic-philosophy-oskari-kuusela/
https://ebookmass.com/product/the-oxford-handbook-of-
experimental-syntax-jon-sprouse/
Precision Engineering 54 (2018) 81–90
Precision Engineering
journal homepage: www.elsevier.com/locate/precision
A R T I C LE I N FO A B S T R A C T
Keywords: Mechanical bearings (i.e., sliding and rolling bearings) are widely used for motion guidance in precision posi-
Pre-sliding/pre-rolling friction tioning stages due to their low cost, large motion range and high off-axis stiffness. They are also finding in-
Bearing creasing use in ultra-precision positioning, e.g., for low-cost and long-range nanopositioning in vacuum en-
Point-to-point positioning vironments. However, mechanical-bearing-guided motion stages suffer from nonlinear pre-motion (i.e., pre-
Disturbance observer
sliding/pre-rolling) friction which adversely affects their precision and motion speed in both tracking and point-
to-point positioning applications. A compliant joint method has recently been proposed for simple, accurate and
robust feedforward compensation of pre-motion friction in tracking motions, with excellent results. This paper
experimentally investigates the influence of the compliant joint method on feedback compensation of pre-motion
friction, which is critical to achieving fast settling in point-to-point positioning. It shows using a model-free (PID)
controller that, for the same feedback gains, the mechanical-bearing-guided motion stage equipped with com-
pliant joints exhibits much more linear closed loop dynamics and higher bandwidth compared to the traditional
motion stage without compliant joints. The compliant-joint-equipped stage also has much faster settling time in
point-to-point positioning experiments for most step motions tested, except for one particular step size where it
settles slower than the traditional mechanical-bearing-guided motion stage due to the compliant joint dynamics.
With the addition of an inverse-model-based disturbance observer to the PID controller, the settling time of the
stage with compliant joints becomes uniformly much faster than the traditional mechanical-bearing-guided
motion stage; its robustness and stability margins are also shown to be superior to those of the traditional
mechanical-bearing-guided motion stage.
1. Introduction PID, P-PI, PI-P, etc.), widely used in practice, often encounter difficul-
ties when trying to overcome the highly varying stiffness of pre-motion
Mechanical bearings (i.e., sliding, and especially, rolling bearings) friction [2,4,5,13–16]. For example, during point-to-point positioning,
are the most cost-effective bearing types used in motion stages [1,2]. the stage is commanded to travel to and settle within a pre-specified
Accordingly, mechanical-bearing-guided motion stages are widely used vicinity (window) of a target position as fast as possible. Pre-motion
in precision applications due to their large motion range, high off-axis friction dominates as the stage approaches its target position, leading to
stiffness and excellent in-position stability [1,2]. Mechanical-bearing- very sluggish settling performance [2,4,10,11,13]. Such long settling
guided motion stages are also very attractive as low-cost alternatives to times severely hamper motion speed. Similarly, during tracking appli-
air bearing stages for a wide range of ultra-precision applications. For cations (e.g., circular tracking or triangular scanning), large position
instance, they are currently the only commercially viable option for a errors (glitches) often occur as the feedback controller tries to overcome
growing number of long-range nanopositioning applications that re- pre-motion friction at motion reversals, jeopardizing motion precision
quire vacuum compatibility [2,3]. [8,9,12].
However, mechanical-bearing-guided motion stages experience To deal with these problems, PID-type controllers must have high
nonlinear pre-motion (i.e., pre-sliding/pre-rolling) friction which ad- gains (i.e., high-gain feedback) in order to quickly overcome the large
versely affects their positioning precision and speed [2,4–13]. In the stiffness of pre-motion friction (during settling or motion reversals)
pre-motion regime, friction behaves as a nonlinear spring due to elas- [4,14,15]. However, such high-gain controllers could easily lead to
toplastic deformations and micro-slip of their inherent rolling elements, large overshoots, limit cycles and instabilities due to the rapid and
end seals and wipers [2,4,5,10–12]. PID-type feedback controllers (e.g., nonlinear changes of pre-motion frictional stiffness during transient
∗
Corresponding author.
E-mail address: okwudire@umich.edu (C.E. Okwudire).
https://doi.org/10.1016/j.precisioneng.2018.05.004
Received 31 October 2017; Received in revised form 9 May 2018; Accepted 17 May 2018
Available online 22 May 2018
0141-6359/ © 2018 Elsevier Inc. All rights reserved.
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
motions [2,5,16]. Therefore, model-based friction compensation approach, called the compliant joint method, which is simpler (and
methods are often used to mitigate the undesirable effects of pre-motion more cost-effective) compared to coarse-fine arrangements and VAN.
friction beyond what is achievable using model-free (e.g., PID-type) The paper [29] shows that pre-motion friction can be accurately and
controllers; they can be executed in feedforward or feedback robustly compensated in feedforward using simple friction models by
[5,8–12,17–21]. Feedforward friction compensation can significantly attaching the bearing to the moving table using a joint that is very
improve the tracking performance of mechanical-bearing-guided mo- compliant in the motion direction. Superior performance and robust-
tion stages when the friction model employed is sufficiently accurate ness of a simple feedforward friction compensation scheme combined
[8,12]. However, it is not effective in solving the slow settling problem with the compliant joint method has been experimentally demonstrated
caused by pre-motion friction during point-to-point positioning, be- through circular tracking motions with different radii and velocities
cause feedforward friction compensation depends on the desired velo- [29,30]. However, the influence of the compliant joint method on
city to predict and preemptively cancel out friction [2,10]. When a feedback compensation of pre-motion friction of mechanical-bearing-
stage is trying to settle to a target position, the desired velocity is often guided motion stage, which is critical to improving settling perfor-
zero, even though the actual velocity is not. Model-based feedback mance in point-to-point positioning, has not been explored.
friction compensation approaches make use of the actual states (e.g., The key contribution of this paper is in carrying out a rigorous ex-
position and velocity) of the system to improve disturbance rejection, perimental investigation into the effects of the compliant joint method
using disturbance observer [8,19,20], gain scheduling controllers on feedback compensation of pre-motion friction, hence its ability to
[4,21], friction observers [17], etc. Because pre-motion friction is ex- improve settling performance of mechanical-bearing-guided motion
tremely nonlinear and variable (i.e., difficult to model accurately), stage in point-to-point positioning. Specifically, after a brief overview
stability and robustness issues often occur when using feedback friction of the compliant joint method and description of the experimental set-
compensation methods, thus limiting their practicality [2,5]. Adaptive up in Section 2, the paper:
control can enable a model-based feedback friction compensation
controller to handle varying friction dynamics [18,19]. However, such 1) Shows in Section 3 using a PID controller that, for the same feedback
approaches often suffer from convergence issues related to poor per- gains, a mechanical-bearing-guided motion stage equipped with
sistence of excitation from smooth reference commands commonly used compliant joints (i.e., compliant stage) exhibits much more linear
in precision applications [5]. closed loop dynamics and higher bandwidth compared to a tradi-
Apart from the abovementioned control-based friction compensa- tional mechanical-bearing-guided motion stage without compliant
tion approaches, the undesirable effects of pre-motion friction can also joints (i.e., rigid stage). Moreover, the compliant stage settles much
be mitigated through design-based methods. This often involves me- faster in point-to-point positioning for most step motions tested,
chanical modification of the traditional mechanical-bearing-guided except for one particular step size during which it settles much
motion stages. For example, a coarse-fine arrangement, where a “fine” slower than the rigid stage due to the dynamics of the compliant
flexure-bearing-guided motion stage is mounted on a “coarse” me- joints.
chanical-bearing-guided motion stage, is sometimes used to improve 2) Demonstrates in Section 4 that with the addition of an inverse-
the precision and speed of mechanical-bearing-guided motion stages model-based disturbance observer (DOB) to the PID controller, the
[22–24]. However, this arrangement makes the system more complex, compliant stage achieves uniformly much faster settling time than
bulky and expensive due to the additional physical components (e.g., the rigid stage; its robustness and stability margins are also shown to
extra sensors, actuators and control hardware) [25,26]. Alternatively, be superior to those of the rigid stage.
Dong et al. [2] have proposed an approach, called vibration assisted This is followed by conclusions and future work in Section 5.
nanopositioning (VAN), to mitigate the slow settling problem of me-
chanical-bearing-guided motion stage using high frequency vibration
(aka dither). Unlike traditional dithering techniques which could jeo- 2. Overview of compliant joint method and experimental set-up
pardize the stage's precision by directly vibrating the stage or guideway
[27,28], VAN is able to improve the settling performance of mechan- 2.1. Compliant joint method
ical-bearing-guided motion stage without vibrating the stage, thus
maintaining high precision. However, the need for additional costly Friction behavior can be divided into two regimes: macro- and
actuators (e.g., piezo actuators and voltage amplifiers) to induce vi- micro-displacement regimes [2,4–7,10–12,29,30]. They are sometimes
brations could limit the practicality of the VAN approach. A recent also referred to as the gross motion and pre-motion friction regimes,
paper [29] by the present authors has proposed a new design-based respectively, where “motion” implies sliding and/or rolling [10–12]. In
the gross motion regime, friction is mainly a function of the relative
Fig. 1. (a) Schematic of a mechanical-bearing-guided motion stage with the bearing: (a) rigidly attached to the table (i.e., rigid stage), and (c) attached to the table
using the compliant joint (i.e., compliant stage); (b) and (d) are equivalent spring models of (a) and (c), respectively.
82
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
velocity between the two objects moving against each other [5,7]. On experimentally identified using the two popular pre-motion friction
the other hand, in the pre-motion regime, friction is primarily a func- models, the Dahl and generalized Maxwell-slip (GMS) models, and the
tion of displacement rather than velocity, hence could be represented as results have been reported in Ref. [29].
a nonlinear spring [2,4–7,10–12,29]. Fig. 1 shows the schematic of a A compliant joint prototype designed in Ref. [29] is used to attach
traditional mechanical-bearing-guided motion stage (i.e., rigid stage) in each bearing of the mechanical-bearing-guided motion stage to the
which pre-motion friction is modeled as an equivalent spring of stiffness table. Based on the requirement that kj is very small (see Section 2.1),
kf connecting the table to ground [29]. At the start of motion or motion the compliant joint must have orders of magnitude less stiffness (in x
reversal, kf is very large; but as more driving force is applied to coun- direction) than the initial large stiffness of pre-motion friction experi-
teract friction, kf rapidly reduces and eventually becomes zero, allowing enced by the bearing. Moreover, it must also maintain the same order of
gross motion of the stage [4–7,10–12]. This leads to a highly nonlinear magnitude of off-axis stiffness as the bearing so as not to unduly com-
time-varying system which is very hard to control [2,4,5,13–16]. In promise rigidity of the stage in the non-motion (i.e., y and z) directions.
particular, the high initial stiffness of pre-motion friction leads to large A flexure is adopted for the design due to its non-contact and friction-
position errors (e.g., quadrant glitches) and slow settling behavior of free nature. To reduce the stiffness in the motion direction, the positive-
mechanical-bearing-guided motion stage, as discussed in Section 1. stiffness flexure is combined in parallel with a negative-stiffness me-
Fig. 1 (c) shows the concept of the compliant joint method (pro- chanism composed of repelling permanent magnets (PMs) to keep the
posed by authors in Ref. [29]). Rather than being rigidly attached to the net stiffness positive but smaller than that of the flexure alone (see
moving table, the bearing of a mechanical-bearing-guided motion stage Fig. 2 (b)). Fig. 2 (c) shows the manufactured compliant joint prototype.
is attached using a joint of stiffness kj in the motion direction. Ac- Details of the compliant joint design and manufacture can be found in
cordingly, the stage with the compliant joint (i.e., compliant stage) can Ref. [29]. Note that the particular design of the compliant joint pro-
be modeled as a series combination of kf and kj, with combined stiffness totype shown in Fig. 2 may introduce Abbe errors to the ultra-precision
kc = kfkj/(kf + kj). Note that the bearing mass is neglected in the spring mechanical-bearing-guided motion stage. These issues are not ad-
model for the following quasi-static stiffness analysis. A very small kj dressed in this paper since they are not relevant to evaluating the
dominates the combined stiffness felt by the stage when kf is very large performance of the proposed method. However, they can be resolved by
in the pre-motion regime (i.e. kc ⟶ kj even when kf ⟶ ∞). As a result, standard precision engineering approaches, such as embedding the
the sensitivity of the combined stiffness kc to variations in kf is given by compliant joint into a pocket in the moving table to reduce the overall
2
profile of the stage.
∂kc η ⎞ kj
= ⎛⎜ ⎟ ; where η=
∂kf ⎝ 1 + η⎠ kf (1) 3. Effects of compliant joint method on model-free (PID)
controlled mechanical-bearing-guided motion stage
Note that if kj ≪ kf, η → 0, and the sensitivity of kc to variations in kf
becomes very small. This indicates that if model-based friction com- 3.1. Frequency-domain evaluation
pensation is performed using kc instead of kf, the adverse effects of
errors in kf (due to low-fidelity friction modeling or variations of fric- Fig. 4 (a) shows the measured frequency response functions (FRFs)
tion) are diminished. In the context of feedforward compensation of of the plant dynamics for rigid and compliant stages (i.e., from input
pre-motion friction, we experimentally demonstrated the compliant force to output displacement). Each FRF is obtained by applying con-
joint method's ability to deliver large reductions of tracking errors, stant-amplitude sinusoidal motor current commands with varying fre-
using low-fidelity friction models, with remarkable robustness in the quencies to the linear motor and recording the corresponding output
presence of up to 50% variations in kf [29,30]. displacement magnitudes of the table position using the linear encoder.
In this paper, we seek to experimentally test the hypothesis that the The motor currents are scaled by a force constant of 11 N/A to convert
compliant joint method also improves the performance and robustness them to equivalent motor forces. It is well known that, in the presence
of feedback friction compensation, hence helps to reduce settling times of pre-motion friction, FRFs of the system can vary significantly, de-
in point-to-point positioning. The rationale for this hypothesis is that pending on excitation amplitudes [6,10–12,31]. Therefore, to demon-
the compliant joint method replaces the very stiff and variable kf with strate the effects of pre-motion friction on rigid and compliant stages,
the much softer and less variable kc, making it much easier for a the FRFs are generated using input force amplitudes ranging from
feedback controller with conservatively-tuned gains to deliver high 0.11 N to 16.5 N (via motor current amplitudes varying from 0.01 A to
performance and robustness. 1.5 A, respectively). When the input amplitude is very small, motion of
the rigid stage is also very small due to the initial large stiffness of pre-
2.2. Experimental set-up: an ultra-precision mechanical-bearing-guided motion friction after each motion reversal. As a result, the measured
motion stage with compliant joints plant dynamics behaves as a standard spring-mass system (as shown in
Fig. 1 (b)) with one dominant low-frequency resonance induced by the
For the purposes of experimentally testing the above-discussed hy- large frictional stiffness. As the input amplitude increases, the frictional
pothesis, an in-house-built ultra-precision mechanical-bearing-guided stiffness gradually decreases, causing the friction-induced resonance to
motion stage equipped with compliant joint prototypes, shown in Fig. 2 shift to lower frequencies [6,10–12], indicating a gradual transition
(a), is utilized. The stage has 1.4 kg moving mass and 40 mm travel from pre-motion to gross motion friction. In the meantime, the motion
range. It is guided by a pair of high-rigidity pre-loaded linear ball of the rigid stage also increases as indicated by larger DC (low-fre-
bearings with end seals (THK, SR-15SB), riding on a super-precision quency) gains of the FRF. Eventually, at the highest input amplitudes,
grade rail, lubricated using grease (THK, AFB-LF). An air core linear the rigid stage experiences pure gross motion and the friction-induced
motor (Aerotech, BLMUC-95), powered by a linear amplifier (Trust low-frequency resonance disappears from the measured FRFs. Notice
Automation, TA-310), is employed to drive the stage. The table position that the high-frequency dynamics of the rigid stage are also greatly
is measured using a linear encoder system (Renishaw, T1000 read head affected by the variations of pre-motion friction. For example, the
and RGSZ20 scale) with post interpolation resolution of 4.88 nm. A magnitude/phase and resonance frequency of the mode around 400 Hz
model of the friction dynamics of the stage is not required for the ex- vary significantly due to its proximity to, and interactions with, the
perimental study presented in this paper. However, to give the reader friction-induced resonance of the stage. Therefore, in agreement with
some sense of the friction behavior, Fig. 3 shows the nonlinear stiffness the literature [6,10–12], the plant dynamics of the rigid stage show
characteristics of pre-motion friction measured using the ultra-precision large variations (nonlinearity) in both low and high frequency regions
motion stage shown in Fig. 2. The friction dynamics have also been with changing input force magnitudes, posing significant challenges for
83
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
Fig. 2. (a) An in-house built ultra-precision mechanical-bearing-guided motion stage equipped with compliant joints, (b) CAD schematic of the compliant joint and
(c) manufactured compliant joint. (Note: the linear encoder is mounted on the distal side of the table and the fixtures can be used to deactivate compliant joints).
much higher, compared to those of the rigid stage. As the input force
amplitude increases, the FRFs of the compliant stage experience similar
transitions as the rigid stage, however, with less variations (of the
magnitude/phase and friction-induced resonance frequencies). More-
over, the higher-frequency dynamics of the compliant stage (e.g., the
mode at 400 Hz) are virtually unaffected by the changing frictional
stiffness, since the friction-induced resonance of the compliant stage
occurs at much lower frequencies. As a result, the compliant stage
shows smaller net stiffness and variations (nonlinearity) in the mea-
sured plant FRFs, and the variations mainly occur at low frequencies
(i.e., less than 100 Hz).
An industry-standard linear-time-invariant (LTI) PID controller is
used to control the stage, as an example of the model-free feedback
controller. It is implemented using a real-time control board (dSPACE,
DS1007) running at 10 kHz sampling frequency. The PID controller is
tuned to 180 Hz closed loop bandwidth using loop shaping, based on an
Fig. 3. Nonlinear stiffness behavior of pre-motion friction measured on the FRF of the rigid stage in the gross motion regime (i.e., using the 1.5A-
ultra-precision stage in Fig. 2. input FRF of the rigid stage in Fig. 4 (a)). The resultant P, I and D gains
are 0.72 N/μm, 138 N/(μm·s) and 8.86 × 10−4 N s/μm, respectively.
feedback control [2,4,5,13–16]. The exact same PID gains are used for controlling the compliant stage. Fig. 4
Since the compliant joint stiffness is designed to be two orders of (b) shows the measured closed loop dynamics (from desired to actual
magnitude smaller than the initial large stiffness of pre-motion friction position) of the rigid and compliant stages using the PID controller.
[29], the combined spring stiffness (in series) experienced by the Because the controller is tuned based on the gross-motion-regime plant
compliant stage at very small input amplitudes is dominated by the dynamics, it encounters difficulties in overcoming the large frictional
compliant joint. This can be seen from Fig. 4 (a); with the same input stiffness of the rigid stage in the pre-motion regime. This leads to sig-
amplitudes applied to the stage, the friction-induced resonance fre- nificant drops of the closed loop FRF magnitudes (below 0 dB) at low
quencies of the compliant stage are much lower, and its DC gains are frequency regions when the input amplitude is small (and pre-motion
Fig. 4. Frequency response functions (FRFs) of: (a) plant, and (b) closed loop dynamics for rigid and compliant stages using different input amplitudes. The rigid
stage exhibits large stiffness, and variations (nonlinearity) in FRF's magnitude/phase and resonance frequencies compared to the compliant stage.
84
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
Fig. 5. Calculated (a) closed loop bandwidth (based on −3 dB criterion), (b) gain and (c) phase margins of rigid and compliant stages with the same PID controller.
friction dominates). As a result, the −3 dB closed loop bandwidth of the most often improved using feedforward friction compensation, for
rigid stage is reduced from its original value of 180 Hz in the gross which the benefits of the compliant joint have already been demon-
motion regime to less than 10 Hz in the pre-motion regime as shown in strated [29,30]. Therefore, it is of particular interest here to evaluate
Fig. 5 (a). In other words, when the stage starts from rest, e.g., during settling times in point-to-point positioning, for which feedback control
motion reversals in tracking motions, or after overshoots in the vicinity is critical. To this end, point-to-point motion profiles consisting of 16
of a target position in point-to-point positioning, the closed loop consecutive steps (8 steps forward and 8 steps backward), of various
bandwidth is reduced to less than 10 Hz. This causes large position step sizes ranging from 50 nm to 5 mm are tested on rigid and com-
errors and sluggish settling performance, as discussed in Section 1. To pliant stages. Each profile of 16 steps is repeated thrice and, in each
overcome this shortcoming, PID gains must be increased to improve the trial, the time taken for each step to settle into a ± 25 nm window is
closed loop bandwidth in the pre-motion regime [2,4,14,15]. However, evaluated. The maximum acceleration/velocity and duration of each
such high-gain controller makes the rigid stage to be prone to in- step size command are summarized in Table 1.
stability, sensitivity to sensor noise, integrator wind-up, chattering and Fig. 6 summarizes the mean settling times, together with the cor-
limit cycles, all of which are very detrimental to its performance [5,16]. responding one-standard-deviation band, of rigid and compliant stages
Compared to the rigid stage, the same PID controller applied to the into a ± 25 nm window during point-to-point motions of different step
compliant stage experiences much less difficulties and performance sizes (48 instances for each step size). For all but one of the step sizes
variations between the pre-motion and gross motion regimes of friction evaluated, the compliant stage exhibits much faster and robust settling
(see Fig. 4 (b)). This is because the nonlinearity in the plant dynamics of than the rigid stage. The exception is the 50 μm step size for which the
the compliant stage mainly occurs at low frequency regions; in the compliant stage settles much slower and is much less robust than the
presence of the feedback controller, the nonlinearity is effectively rigid stage.
suppressed through a phenomenon known as “feedback linearization” To illustrate the reason behind this discrepancy, Fig. 7 compares the
[32]. In other words, since the equivalent frictional stiffness experi- typical settling performance of rigid and compliant stages during point-
enced by the compliant stage is much smaller and less variable, the to-point motions with the 5 μm and 50 μm step sizes. During the first
feedback controller can better overcome and regulate the disturbance step of the 5 μm case (as highlighted in the subplot), the rigid stage
force originating from it. Therefore, a conservatively-tuned PID con- takes 47.8 ms to settle, which is very long relative to the duration of the
troller (from the stand-point of the rigid stage) acts as a “high-gain” step command (i.e., 17.7 ms); however, the compliant stage settles
feedback controller for the compliant stage, leading to more-consistent within 10.5 ms (i.e., 78% faster). Similarly, during the eighth step, the
and improved closed loop bandwidth for both pre-motion and gross compliant stage takes 10.2 ms to settle, i.e., 88% faster than the rigid
motion regimes of friction, as shown in Fig. 5 (a). stage (which takes 85 ms to settle). Similar improvements in settling
Fig. 5 also shows the calculated gain and phase margins of the rigid time are seen for all 16 steps (8 steps forward and 8 steps backward)
and compliant stages with different input amplitudes. It is assumed that such that the mean settling time of the compliant stage (i.e., 10.3 ms) is
each measured FRF of the stage (at a given input amplitude) represents 85.1% faster than that of the rigid stage (i.e., 69.1 ms). Fig. 7 (b)
an LTI system. This is essentially equivalent to linearizing the nonlinear compares the settling response of rigid and compliant stages for the
plant at different operating points (i.e., different pre-motion frictional problematic 50 μm step size. During the first step, the rigid stage takes
stiffness values). Although local stability of the linearized systems does 237 ms to settle, while the compliant stage only takes 43.6 ms to settle,
not guarantee the global stability of the nonlinear system, the stability leading to 81.6% reduction in settling time. However, it is observed that
of the individual LTI systems can be used as an indicator of the stability the settling time of the compliant stage gradually increases as the stage
of rigid and compliant stages at various operating points. For example, steps in the same direction. By the eighth step (as highlighted in the
it is observed that stability margins of both stages are always positive, subplot), the settling time of the compliant stage has increased to
showing that each individual LTI system is indeed stable. Moreover, the 453 ms, which is even longer than that of the rigid stage (i.e., 214 ms).
compliant stage maintains similar or even slightly better stability This drastic change in settling performance leads to the overall longer
margins (robustness) while achieving much higher closed loop band- mean settling time and larger one-standard-deviation band during the
width, compared to the rigid stage. This is mainly due to less interaction
between the friction-induced resonance and the higher-frequency dy-
Table 1
namics of the compliant stage compared to the rigid stage (see Fig. 4 Maximum acceleration/velocity and duration of each step size command.
(a)).
Step size Max. acceleration Max. velocity Duration per step
85
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
Fig. 8. Schematic of table and bearing motions of the compliant stage during
settling, explaining the reason for long settling time when dxb/dt > 0 com-
pared to when dxb/dt ≈ 0.
86
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
87
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
Fig. 11. Frequency response functions (FRFs) of (a) augmented plant and (b) closed loop dynamics for rigid and compliant stages with PID controller combined with
DOB (of 200 Hz cutoff frequency) using different input amplitudes. The rigid stage exhibits large stiffness, and variations (nonlinearity) in FRF's magnitude/phase
and resonance frequencies compared to the compliant stage; the bandwidth of the rigid stage's DOB is reduced to 100 Hz to guarantee stability (as discussed in
Section 4.3).
Fig. 12. Calculated (a) closed loop bandwidth (based on −3 dB criterion), (b) gain and (c) phase margins of rigid and compliant stages with PID controller combined
with DOB. The DOB of 200 Hz bandwidth is stable on the compliant stage but unstable on the rigid stage; the bandwidth of the rigid stage's DOB is reduced to 100 Hz
to guarantee stability (as discussed in Section 4.3).
4.3. Time-domain evaluation using point-to-point positioning tests 3.2. The same PID controller as in Section 3 is used on both stages.
However, while the DOB tuned in Section 4.1 works for the compliant
Point-to-point positioning tests are carried out on the rigid and stage, it leads to instability when used on the rigid stage, as predicted
compliant stages using the same motion profiles introduced in Section by the stability margin plots of Fig. 12. Fig. 13 shows an example of the
88
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
and 29.5 ms to settle in the first and eighth steps while the compliant
stage settles within 7 ms and 6.7 ms, respectively. During point-to-point
motions with the 50 μm step size, the rigid stage takes 39.5 ms and 56.1
ms to settle in the first and eighth steps, respectively, while the com-
pliant stage takes 13.5 ms and 13.3 ms to settle in the first and eight
steps. This shows that the undesirable effects of the bearing motion
during settling is effectively attenuated when the DOB is implemented
in the control scheme. Although the bearing still moves significantly
during settling for the 50 μm case, the DOB is able to effectively esti-
mate and cancel out the disturbance force originating from its motion.
This significantly reduces the workload of the PID controller such that it
Fig. 13. Typical positioning performance of the rigid stage with DOB during
can quickly move the compliant stage to the target position.
point-to-point motions with the 5 mm step size. The stage suffers from severe
oscillations during settling when the cutoff frequency of the DOB is set to
150 Hz and 200 Hz.
5. Conclusions and future work
89
X. Dong, C.E. Okwudire Precision Engineering 54 (2018) 81–90
Fig. 15. Typical positioning performance of rigid and compliant stages with DOB into a ± 25 nm window during point-to-point motions with (a) 5 μm and (b) 50 μm
step sizes. The first and eighth steps in the forward direction are highlighted in the subplots.
Acknowledgements [17] Canudas de Wit C, Olsson H, Åström K, Lischinsky P. A new model for control of
systems with friction. IEEE Trans Automat Contr 1995;40(3):419–25.
[18] Canudas de Wit C, Lischinsky P. Adaptive friction compensation with partially
The authors are grateful to Aerotech Inc. for donating some of the known dynamic friction model. Int J Adapt Contr Signal Process 1997;11(1):65–80.
key components used in building the experimental set-up. This work is [19] Olsson H, Åström K, Canudas de Wit C, Gäfvert M, Lischinsky P. Friction models and
funded by National Science Foundation Award CMMI #1562297: friction compensation. Eur J Contr 1998;4(3):176–95.
[20] Ishikawa J, Tomizuka M. Pivot friction compensation using an accelerometer and a
Vibration Assisted Nanopositioning: An Enabler of Low-cost, High- disturbance observer for hard disk drives. IEEE ASME Trans Mechatron
throughput Nanotech Processes. 1998;3(3):194–201.
[21] Cheng G, Hu J. An observer-based mode switching control scheme for improved
position regulation in servomotors. IEEE Trans Contr Syst Technol
References 2014;22(5):1883–91.
[22] Numasato H, Tomizuka M. Settling control and performance of a dual-actuator
[1] Altintas Y, Verl A, Brecher C, Uriarte L, Pritschow G. Machine tool feed drives. CIRP system for hard disk drives. IEEE ASME Trans Mechatron 2003;8(4):431–8.
Ann - Manuf Technol 2011;60(2):779–96. [23] Elfizy A, Bone G, Elbestawi M. Design and control of a dual-stage feed drive. Int J
[2] Dong X, Yoon D, Okwudire C. A novel approach for mitigating the effects of pre- Mach Tool Manufact 2005;45(2):153–65.
rolling/pre-sliding friction on the settling time of rolling bearing nanopositioning [24] Fung R, Hsu Y, Huang M. System identification of a dual-stage xy precision posi-
stages using high frequency vibration. Precis Eng 2017;47:375–88. tioning table. Precis Eng 2009;33(1):71–80.
[3] Sitti M. Survey of nanomanipulation systems. Proceeding of the 1st IEEE conference [25] Hao G, Kong X. Novel xy compliant parallel manipulators for large displacement.
on nanotechnology. 2001. p. 75–80. Proceedings of the ASME international design engineering technical conference &
[4] Futami S, Furutani A, Yoshida S. Nanometer positioning and its micro-dynamics. computers and information in engineering conference, vol. 44. 2010. p. 1–11.
Nanotechnology 1990;1(1):31–7. [26] Awtar S, Parmar G. Design of a large range xy nanopositioning system. J Mech
[5] Armstrong-Hélouvry B, Dupont P, Canudas de Wit C. A survey of models, analysis Robot 2013;5(2). 021008–1 to 021008–10.
tools and compensation methods for the control of machines with friction. [27] Engel T, Lechler A, Verl A. Sliding bearing with adjustable friction properties. CIRP
Automatica 1994;30(7):1083–138. Ann - Manuf Technol 2016;65(1):353–6.
[6] Otsuka J, Masuda T. The influence of nonlinear spring behavior of rolling elements [28] Tanaka T, Oiwa T, Syamsul H. Positioning behavior resulting from the application
on ultraprecision positioning control systems. Nanotechnology 1998;9(2):85–92. of ultrasonic oscillation to a linear motion ball bearing during step motion. Precis
[7] Al-Bender F, Swevers J. Characterization of friction force dynamics. IEEE Contr Syst Eng 2018;51:362–72.
Mag 2008;28(6):64–81. [29] Dong X, Liu X, Yoon D, Okwudire C. Simple and robust feedforward compensation
[8] Jamaludin Z, Van Brussel H, Pipeleers G, Swevers J. Accurate motion control of xy of quadrant glitches using a compliant joint. CIRP Ann - Manuf Technol
high-speed linear drives using friction model feedforward and cutting forces esti- 2017;66(1):353–6.
mation. CIRP Ann - Manuf Technol 2008;57(1):403–6. [30] Dong X, Okwudire C. Detailed experimental evaluation of the compliant joint
[9] Fukada S, Fang B, Shigeno A. Experimental analysis and simulation of nonlinear method for feedforward compensation of pre-motion friction. 32th annual meeting
microscopic behavior of ball screw mechanism for ultra-precision positioning. of the American society for precision engineering. 2017.
Precis Eng 2011;35(4):650–68. [31] Helmick D, Messner W. Describing function analysis of dahl model friction.
[10] Bucci B, Vipperman J, Cole D, Ludwick S. Evaluation of a servo settling algorithm. Proceedings of the American control conference. vol. 0(5). 2009. p. 814–9.
Precis Eng 2013;37(1):10–22. [32] Leang K, Devasia S. Feedback-linearized inverse feedforward for creep, hysteresis,
[11] Maeda Y, Iwasaki M. Rolling friction model-based analyses and compensation for and vibration compensation in AFM piezoactuators. IEEE Trans Contr Syst Technol
slow settling response in precise positioning. IEEE Trans Ind Electron 2007;15(5):927–35.
2013;60(12):5841–53. [33] Umeno T, Hori Y. Robust speed control of DC servomotors using modern two de-
[12] Yoon J, Trumper D. Friction modeling, identification, and compensation based on grees-of-freedom controller design. IEEE Trans Ind Electron 1991;38(5):363–8.
friction hysteresis and Dahl resonance. Mechatronics 2014;24(6):734–41. [34] Kempf C, Kobayashi S. Disturbance observer and feedforward design for a high-
[13] Ruderman M, Iwasaki M. Analysis of linear feedback position control in presence of speed direct-drive positioning table. IEEE Trans Contr Syst Technol
presliding friction. IEEJ J Ind Appl 2016;5(2):61–8. 1999;7(5):513–26.
[14] Chang S, Wu S, Hu Y. Submicrometer overshoot control of rapid and precise posi- [35] Schrijver E, van Dijk J. Disturbance observers for rigid mechanical systems:
tioning. Precis Eng 1997;20(3):161–70. equivalence, stability, and design. J Dyn Syst Meas Contr 2002;124(4):539.
[15] Lin T, Pan Y, Hsieh C. Precision-limit positioning of direct drive systems with the [36] Jamaludin Z, Van Brussel H, Swevers J. Friction compensation of a XY feed table
existence of friction. Contr Eng Pract 2003;11(3):233–44. using friction model-based feedforward and an inverse-model-based disturbance
[16] Sato K, Nakamoto K, Shimokohbe A. Practical control of precision positioning observer. IEEE Trans Ind Electron 2009;56(10):2–6.
mechanism with friction. Precis Eng 2004;28(4):426–34.
90
Another random document with
no related content on Scribd:
mountains are so high that the mea nona ua aina la, o
stars [42]appear on them, and Namakaokahai, he ’lii wahine, a
there are very few people living he mau kaikunane kona eha, o
on it. The owner of the land is Kanemoe, o Kaneikaapua, o
Namakaokahai, a chiefess, and Leapua, a me Kahaumana. Elua
she has four brothers: Kanemoe, kauwa, o Upoho, a me
Kaneikaapua, Leapua and Haapuainanea. O ke kiai o ka
Kahaumana. She has two aina, o Moela, he ilio. Ekolu
servants, Upoho and manu, o Manuea, o Kiwaha, a
Haapuainanea. Those who me Halulu. Oia na kanaka o ia
guard and watch over the land aina, aohe nui, no ka mea, he
are Moela, a dog, and three pau i ka ai ia e ke ’kua.”
birds, Manuea, Kiwaha and
Halulu. These are all the people
who live on the land; there are
not many, because the people
are devoured by the ghosts.”
After all the various things had A pau na mea a pau loa i ka loaa
been mastered by Aukelenuiaiku ia Aukelenuiaiku, ia wa, kau ae
he then climbed onto the back of la o Aukelenuiaiku i luna o ka
the lizard and was lifted up out of moo, hapai ae la ia ia
the pit; and the lizard again Aukelenuiaiku a kau i luna o ka
disappeared down the pit. lua, a hoi aku la ka moo i lalo o
Aukelenuiaiku then took up the ka lua. A hoi aku la o
box that contained his god Aukelenuiaiku, lalau iho la ia i ka
together with his club and carried pahu o ke ’kua ona, a me ka
them as he returned to the laau, a hii ae la, a hoi aku la i ka
house. When Aukelenuiaiku hale. Ia Aukelenuiaiku i hiki ai i
reached the house his father and ka hale, uwe mai la kona
all the chiefs wept for joy. [44] makuakane a me na ’lii a pau
loa. [45]
At the end of the first day and on A hala ka la ana i hoakaka ai,
the approach of the second, the hiki mai ka lua o ka la, o ia kana
day on which he had said they la i olelo ai e ku i ka aina o
were going to arrive at Holaniku, Holaniku. I ke kakahiaka nui, ike
very early that morning they first mua ia mai la ke kuahiwi o ka
saw the peaks of the mountains, aina, a awakea, ku lakou i ka
and by noon of that same day aina. Ma keia ku ana, lele aku la
they reached the land. As soon na kanaka i uka, loaa ka ai, ka
as the ship touched land the wai, ka ia, ka awa a me na mea
men went ashore where they e ae, o ka nui o ko lakou
found food, water, meat, awa manawa i laila, eha po, eha ao.
and various other things. They Kau lakou i ka moku a holo aku
stayed on the land for four days la.
and four nights, when they again Ma keia holo ana, eha o lakou
boarded their ship and set sail. malama i ka moana, a pau ia
After sailing for four months, mau malama eha, hai aku o
Aukelenuiaiku said to his Aukelenuiaiku ia lakou: “Apopo
brothers: “Tomorrow we will kakou ku i ka aina.” A lohe na
reach land.” When his brothers kaikuaana, olelo mai lakou,
heard this, they said: “You are “wahahee oe;” aka, aohe mea
deceiving us.” But there was nana e hoole mai o
none of them who could deny Aukelenuiaiku ma kona mau ano
the fact, for Aukelenuiaiku ike a me ke akamai, ua ae no ka
showed that he knew what he poe holo moku a pau loa, a o na
was talking about; so the sailing kaikuaana wale no ka poe hoole,
masters all admitted that no ko lakou opu inoino ia
Aukelenuiaiku was correct. But Aukelenuiaiku.
the brothers being bitter against
Aukelenuiaiku, refused to believe
him.
CHAPTER V. MOKUNA V.