Printer Friendly

In-mold multivariate sensing of colored polystyrene.

INTRODUCTION

Injection molding manufacturers are constantly striving for higher part quality and less waste. Mechanical properties can be improved by fibers, price can be reduced by using talcum powder, electrical conductivity can be improved using carbon, but the most common additive is color [1]. Whenever an additive is used in a plastics manufacturing, the rheological properties of the resin are influenced by the additives [2]. Customers commonly request different colors and surface finishes, and the aesthetic qualities of the molded components have been found to be related holding pressure and filling rate [3]. To remain competitive, plastics manufacturers are utilizing fully automated systems for fault diagnosis, quality control, and material handling [4]. Advances in machine control have increased the repeatability of the injection molding process drastically [5-7]. Agrawal et al. categorized controlled variables into all-phase control, phase-dependent control, and cycle-to-cycle control to improve overall process control [5]. Schulz described the advances in machining and mold building technology [6]. Giboz et al. described how laser micromachining and electrical discharge machining allowed for manufacturing of molds with repeatable features on the micro scale [7]. Machine control in injection molding is an important factor; however, there is still difficulty in control of the transient and distributed state of the polymer melt. There are four key states known to govern the quality of injection molded parts: melt temperature, pressure, velocity, and apparent viscosity [8-12].

Highly responsive infrared (IR) temperature sensors are commonly found in polymer processing [13]. There is a fundamental issue when measuring IR melt temperature, which is varying emissivity. Polymer emissivity can vary with material and color. Prior research found similar polymers with different colors had emissivities varying between 0.74 and 0.92 [14]. Differing polymer emissivity can cause variations in the power and wavelength distribution of the emitted radiation. For this reason, polymer melts having the same melt temperature may be detected as having different melt temperatures via IR pyrometry. To investigate potential issues associated with color, high impact polystyrene (HIPS) was compounded with multiple color additives and experimentally investigated with an instrumented injection mold.

An in-mold multivariate sensor (MVS) has been developed to measure melt pressure, IR melt temperature, mold temperature, and by mechanistic analysis provide an estimate for velocity and viscosity. This MVS design is shown in Fig. 1. Temperature is measured using a HMS Z11-F5.5 IR thermopile (Heimann Sensor GmbH, Dresden, Germany). Pressure is measured using a piezoelectric (PZT) ceramic ring with an 8 mm outer diameter, 4 mm inner diameter, and 0.5 mm thickness made of APC-841 material (Mackeyville, Pennsylvania). The PZT is electrically insulated and centered using custom polyphenylene sulfide (PPS) washers. A zinc selenide (ZnSe) lens is cylindrically ground to fit in the sensor head with a tight fit to transmit IR radiation to the IR detector, transfer stress to the PZT ring, and seal the sensor cavity from the polymer melt. The sensor body is machined using a standard Dynisco style snout blank for comparison with traditional instrumentation.

The determination of polymer morphological properties in addition to temperature and pressure measurement has been shown to directly determine the finished part quality in injection molding [15]. The apparent viscosity is an important melt property, and often used as an indicator of other polymer properties such as molecular weight, modulus, and others. In contrast to a Newtonian fluid, the viscosity of a plastic material depends on temperature and shear rate polymer material [16]. Shear thinning behavior is typical for polymer melt and needs to be considered in polymer processing. The polymer melt velocity is also important for process and quality control. Controlling the velocity profile in the filling phase is critical to part uniformity [17].

This work is motivated to characterize the effect of the color on the MVS measurements of temperature, pressure, velocity, and viscosity. While a robust sensor is desired that is insensitive to color, it is a secondary objective of the ongoing research to try and identify the type of colorant from observed systematic variations in the process signals, if any.

THEORY

Output voltage responses are obtained with the MVS that are indicative of the polymer pressure, polymer irradiation, and mold temperature. After converting these voltages into pressure and temperatures, respectively, the melt velocity and viscosity can then be estimated by mechanistic analysis.

Melt Pressure Measurement

As the polymer melt flows across the MVS, as shown in Fig. 1, the force exerted on the lens by the melt pressure will be transferred onto the PZT ring, which will cause an imposed stress and (by the piezoelectric effect) a proportional charge. The voltage response from the piezoelectric ring, [V.sub.PZT], is described in the following equation:

[V.sub.PZT] = 4 x [g.sub.33] x P x H x [R.sup.2]/[OD.sup.2] - [ID.sup.2] (1)

where [g.sub.33] is the voltage constant determined by the PZT material, H is the ring thickness of 0.5 mm, ID is the ring inner diameter of 4 mm, OD is the ring outer diameter of 8 mm, and R is the lens radius of 1.5 mm. For the design shown in Fig. 1, with H equal to 0.5 mm and a voltage constant of to 25.5 X [10.sup.-3] Vm/N for APC-841 material the voltage response will be 2.39 V/MPa of melt pressure. The sensor is Hush mounted to the surface of the cavity to prevent any losses or abnormalities in the pressure reading that would occur from a gap between the cavity and sensor.

Melt Temperature Measurement

As the polymer melt flows across the sensor window, IR radiation passes through the zinc selenide window, and is collected by the thermopile (TP). The voltage response of the thermopile, [V.sub.TP], described by the following equation:

[V.sub.TP] = k x [epsilon] x [T.sup.n.sub.melt] - [T.sup.n.sub.mold]) (2)

where k is the gain, [epsilon] is the emissivity of the polymer, [T.sub.mclt] and [T.sub.mold] are the absolute temperatures of the melt and the mold, respectively, and n is dependent on the filter and sensor characteristics (equal to 4 for a perfect "black" body and unlimited wavelength range).

Mold Temperature Measurement

The thermopile contains a thermistor to assess the reference temperature of the CMOS IR detector, which must be known to compute the net radiative heat transfer to the thermopile. The 100 k[OMEGA] thermistor resistance is supplied from the manufacturer as a function of temperature to within 0.2% absolute error. A voltage divider circuit was designed to convert the thermistor's output resistance to a voltage. The value of the reference resistor (10 k[OMEGA]) was selected to scale the output voltage to a desired range while also linearizing the thermistor output within the mold coolant temperatures of interest, from 25[degrees]C to 100[degrees]C.

Melt Velocity Estimation

As the polymer melt passes over the lens, irradiation is transmitted to the sensor's thermopile. The amount of the lens exposed to the polymer melt varies as the polymer melt encounters and Fills the portion of the mold cavity adjacent to the lens. Earlier work recognized that the melt velocity could be obtained from the transient melt temperature signal. Specifically, the rise time of an IR pyrometer signal is dependent on the velocity of the polymer melt crossing the sensor's lens as shown in Fig. 2. The underlying reason is that the IR pyrometer senses the irradiation from the polymer melt. If only a portion of the lens is in contact with the polymer, then the amount of emitted radiation is reduced and the temperature sensed by the pyrometer will be proportionally less.

Two simplifying assumptions can be made to estimate the polymer melt velocity. First, the polymer is assumed to have a uniform bulk temperature as it flows past the sensor. Second, the amount of transmitted power is assumed to be proportional to the projected area of the melt. With these two assumptions, the velocity of the polymer melt can be derived from the characteristics of the melt temperature characteristics. Specifically the area, A, of the circular segment with lens radius, R, is equal to:

A = 1/2 [R.sup.2] x ([theta] - sin[theta]) (3)

where [theta] is the fully included angle of the circular segment corresponding to the polymer melt position, x[member of][-R,R], such that [theta] is equal to 2arccos(x/R). The analysis assumes the flow front of the polymer melt is straight with respect to the direction of flow, which is feasible given the small lens radius relative to the cavity width of Fig. 2. The velocity is best estimated when the polymer melt is passing over the center of the lens, which corresponds to the maximum derivative of the area with respect to time and so should provide the greatest signal-to-noise ratio. At this moment, the velocity, [v.sub.x], can be derived as:

[v.sub.x] = [pi]R/2[T.sub.0] x [(dT/dt).sub.max] (4)

where [T.sub.0], is the absolute temperature of the polymer melt as it crosses the center of the lens, and dT/dt is the temperature derivative as a function of time. As the velocity estimate is dependent on the temperature slope, not accounting for variations in emissivity may lead to inaccurate measurements.

Melt Viscosity Estimation

The polymer's shear viscosity is a critical indicator of the polymer morphology, and so is also a determinant of the residual stress distribution and resulting product quality [18, 19]. As the polymer melt continues flowing past the sensor, the sensed melt pressure will increase in proportion to the flow distance. Since the velocity, vx, is the time derivative of the melt position with respect to time, application of the chain rule to the Hele-Shaw equation for viscous flow provides an estimate of the apparent melt viscosity, [mu]:

dP/dx = 12 [mu][v.sub.x]/[H.sup.2] [right arrow] [mu] = [H.sup.2]/12[v.sup.2.sub.x] dP/dt (5)

where H is the known cavity thickness measured from the top of the sensor lens to the opposing mold wall. Ideally, the pressure derivative is calculated immediately as the polymer melt crosses the sensor location. The viscosity estimation is dependent on velocity estimation, which in turn is dependent on IR temperature measurement. The emissivity of the polymer is thus critical to accurate temperature, velocity, and viscosity measurements.

MATERIAL PREPARATION

Dow Styron 478 HIPS was compounded with three color additives (ECM Plastics). The colors used in this experiment were blue, black, and purple, as shown in Table 1. All materials were compounded using a Leistritz twin screw extruder (ZSE 27HP-400) with a screw diameter of 27 mm. The screw configuration was configured for a low shear rate profile so as to prevent damage to the material but still have a sufficient mixing behavior. The temperature profile of the extrusion process is provided in Table 2. To control the material feed of the single components, two independent speed-controlled feeders from Brabender Technologies[R] and Schenck[R] were used to establish the recommended proportion of either 25:1 or 50:1. The extruded material was cooled in a water bath and cut by a Reduction Engineering[R] strand pelletizer. The pellets were then used for viscosity testing in the capillary rheometer (Dynisco LCR7000) with a length to diameter (L:D) ratio of 30:1 mm. The extruder output was set to correspond to an observed output rate of 12.5 kg/h.

The compounding process can have a significant influence on the viscosity of the polymer. The incorporated shear and temperature history will shorten the polymer chain length and lead to a decreasing viscosity. In order to characterize the effect, the material degradation due to compounding has to be considered. Therefore, the viscosity of the neat resin is compared before and after one cycle in the compounding. The viscosity change due to compounding is evidenced by the 0.99 slope of the fit line in Fig. 3, which plots the viscosity of the neat resin as received with the viscosity of the neat resin cycled once through the extruder. The effect of the added colorant on the viscosity of the blends was previously investigated and found to be a statistically significant, with a typical decrease in the viscosity of 1% [20].

INJECTION MOLDING

The instrumented flexural test specimen cavity used in this study is shown at right in Fig. 1. This cavity is part of an ASTM family mold with extensive instrumentation. There are pressure transducers (Priamus 6001 A, Schaffhausen, Switzerland) at the junction between the primary and secondary runner, as well as at the beginning of the flexural test specimen cavity. The MVS was installed at the center of the flexural test specimen cavity. Toward the end of fill, a fiber optic IR pyrometer (Omega OS 1562, Stamford, CT) is installed in place of an ejector pin, and a type N thermocouple (Primaus 4001 A) having an exposed junction. The mold was installed on a Sumitomo SE75DUZ, which is a fully electric 75 ton injection molding machine. Four machine signals, including injection pressure, screw position, screw speed, and screw RPM, were acquired with a Priamus eDAQ 8102A at its maximum acquisition rate of 500 Hz. Data from the two piezoelectric pressure sensors and the type N thermocouple were also obtained. The IR probe and the MVS signals (melt temperature, mold temperature, and melt pressure) were collected using a National Instruments NIUSB6212 at a rate of 10 kHz. The DAQ systems were both triggered using the 24 volt digital mold closed signal that the machine produces.

A 12-run full factorial design of experiments (DOE) was run to measure the effects of the four colors with three different linear melt velocities: 100, 400, and 1000 mm/s. Ten samples were taken for each run. These three velocities were purposefully chosen to correspond to typical, fast, and very fast melt velocities used in injection molding. The screw speed, back pressure, and barrel temperature were maintained for all materials to correspond to a target polymer melt temperature of 220[degrees]C. Using a handheld thermocouple, the melt temperature measured at 217.8[degrees]C. The mold coolant temperature was held constant at 21[degrees]C. The pack pressure used was held constant at 50 MPa for a duration of 7 s, which exceeded the gate freeze time. When changing the type of polymer, at least 30 cycles were produced, after purging, to allow the machine to reach steady-state before the ten samples were collected for each run. The DOE design is shown in Table 3.

Typical injection molding traces are shown in Fig. 4, with the signals acquired from the commercial and MVSs instrumented in the mold as described in Fig. 1. Figure 4a shows three pressure traces. As expected, the pressure is highest in the runner, and lowest at the MVS, which are the closest and farthest pressure sensors to injection, respectively. Figure 4b shows the two IR temperature traces. The MVS experiences a higher temperature spike than the commercial IR sensor, and then displays more of the cooling behavior of the melt; this behavior occurs because the commercial IR sensor can only measure melt temperatures above 80[degrees]C. Finally, Fig. 4c shows the commercial thermocouple and the MVS thermistor. The commercial thermocouple shows a higher temperature spike, but this is mostly because this sensor is in contact with the melt, whereas the thermistor is isolated from the melt by the 5 mm height of the zinc selenide window. The noise found in the MVS thermistor signal is related to the noise found in the MVS IR signal, as they are both obtained using the same sensing element with the IR detector using the thermistor to provide a reference temperature.

A Moldflow simulation of the mold shown in Fig. 1 was performed using 3,974,728 tetrahedral elements including 345,120 elements in the flexural test specimen cavity. The simulated melt viscosity contour plot from this simulation is shown in Fig. 5. The simulation was run using the process settings from Run 1 of the DOE described in Table 3. The apparent melt viscosity found in this simulation was 398 Pa s, which is a useful estimate of the bulk viscosity that would otherwise be unobservable.

RESULTS AND DISCUSSION

There were four concerns regarding color change in this experiment: (a) Will the color change affect the viscosity of the material, and will the viscosity changes be witnessed in pressure measurements? (b) Will the color change affect temperature measurements due to differing emissivities? (c) If the material change alters the temperature measurement, will it disrupt the velocity estimation? (d) Since the viscosity calculation relies on the velocity estimation, which in turn relies on the temperature measurement, how much variation will be seen in the viscosity estimation? The first of these four concerns is investigated in Fig. 6. As expected, the pressure measurements using the MVS were very consistent in all four colors. The cycles chosen to plot were Run 2-5, 5-5, 8-5, and 11-5, correlating to the 400 mm/s runs for white, blue, black, and purple respectively. As previously characterized, the viscosity of the material is very similar for all four materials, so the pressure should be very similar [20].

The temperature measurement, as shown in Fig. 7, is where any variation between materials of varying emissivity would most likely be apparent. Figure 7 shows four different temperature cycles from the same runs as in Fig. 6. Zooming in on the peak temperatures, slight variations between the colors are observed, with a 12[degrees]C difference occurring between the maximum temperature for the neat material and the minimum temperature for the black material. Prior research during preliminary calibration experiments also found a similar behavior. While intuition may suggest that black bodies provide higher emissivity since they more closely emulate an ideal "black" body, the reduced irradiation is due to internal absorption through the thickness of the polymer melt.

Using the data collected, velocity estimates were created using the velocity estimation principle described in Fig. 2 and Eq. 4. Figure 8 compares the polymer melt velocity calculated using just the transient behavior of the MVS' temperature signal and the velocity, v = [DELTA]v/[DELTA]t, determined using the known distance, [DELTA]x, and the known travel time, [DELTA]t, between the commercial in-mold pressure sensor and the MVS. The results show a high correlation of for all samples between the slower two injection velocities, 100 and 400 mm/s, but no differentiation between 400 and 1000 mm/s. The 1000 mm/s velocity results indicate that the response time of the thermopile is insufficient to characterize very fast polymer melt velocities. This result was actually expected since the thermopile was known to have a response time on the order of 2 ms, such that the received power from a polymer melt flowing over the MVS' small lens diameter would be governed by the lag in the thermopile. Still, the MVS velocity estimates were found to be highly insensitive to color with coefficients of determination, [R.sup.2]. for all materials falling in the range of 0.99 [+ or -] 0.004.

There is very little velocity variation with respect to injection molding samples with different color additives. Previous research in viscosity characterization and validation of mixing rules found very little variation in the viscosity with respect to color blends. Figure 9 plots the viscosity estimated on-line with the MVS relative to the Cross-WLF model fit to the characterized viscosity data for the neat material. The on-line viscosity estimates acquired with just the MVS are on the right order of magnitude as the characterized melt viscosity. While shear-thinning behavior of the viscosity is observed, the non-monotonic estimate of the viscosity for the polymer at melt velocities of 400 and 1000 mm/s is erroneous. It is surmised that the behavioral variance is occurring due to the inaccurate estimate of the melt velocity at 1000 mm/s. If this velocity had been correctly estimated by the MVS, the apparent viscosity at the higher velocity would be reduced by a factor of 6.25, according to Eq. 5. One important result from the viscosity data observed in Fig. 9 is that the viscosity variation due to process inconsistencies is greater than the variation between different colors which conforms to previous research and simulated data. Vertical error bars show the uncertainty in viscosity estimations, error in the shear rate calculations were found to be insignificant, so they were removed from Fig. 9.

CONCLUSIONS AND FUTURE WORK

It is important to emphasize that the implemented MVS was designed as a prototype, yet this prototype significantly improves observability in polymer processing. The MVS' melt pressure and mold temperature sensing are already very accurate relative to commercial sensors. The MVS temperature measurements are also very capable compared with commercial IR pyrometers, and provide much broader range for temperature sensing, especially at low temperatures. The 12[degrees]C variation in acquired melt temperatures with respect to different colors is consistent with prior characterization research, and is also observed with commercial IR pyrometers. The use of mechanistic analysis to estimate the melt velocity and viscosity indicates significant potential for process and quality control. The use of the transient temperature data for velocity estimation indicated that the MVS can provide high fidelity estimates for melt velocities up to 400 mm/s, after which the differentiation at higher velocities is lost. The known limitation is the response time of the thermopile.

In order to create a commercially viable MVS, a few changes must be made. First, the IR detector must be able to characterize faster polymer melt velocities. Finding the right IR detector is not trivial. Phototodiodes, such as InGaAs, provide very fast response but limited range compared with the implemented thermopile. Second, to improve the signal-to-noise ratio, which is critical at high resolution and acquisition rates, it is possible to encapsulate all the signal conditioning and data acquisition with the sensor body itself through the use of an application-specific integrated circuit (ASIC). Incorporating the ASIC into the MVS would thus create a compact and robust sensor providing a set of digital output signals for many process states. Indeed, advances in electronic miniaturization are continuing to enable new sensing elements as well as real-time computing within the sensor itself. The next generation MVS may contain a CMOS image sensor for red/green/blue/IR detection as well as a pressure sensor. Such in-mold imaging would obviate the need for fast IR temperature detection since the velocity estimation could be performed directly through consecutive frames of acquired images interfaced directly to next generation of molding machines.

ACKNOWLEDGMENTS

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. The authors would also like to thank Priamus System Technologies, LLC for the use of their sensors and data acquisition system as well as Dynisco Instruments for their support in providing the sensor snout blanks, and ECM Plastics for providing the color additives. A special thanks to Professor Stephen P. Johnston and Gabriel Mendible from The University of Massachusetts--Lowell for running the Moldflow simulations.

REFERENCES

[1.] W. Thongruang, R.J. Spontak, and C.M. Balik, Polymer, 43, 2279 (2002).

[2.] J.M.J. Toonder, F.T.M. Nieuwstadt, and G.D.C. Kuiken, Appl. Sci. Res., 54, 95 (1995).

[3.] F. Pisciotti, A. Boldizar, M. Rigdahl, and I. Arino, Polym. Eng. Sci., 45, 1557 (2005).

[4.] D. Kazmer, S. Johnston, and R. Gao, Int. Polym. Process., 26, 63 (2011).

[5.] A. Agrawal, I. Pandelidis, and M. Pecht, Polym. Eng. Sci., 27, 1345 (1987).

[6.] H. Schulz, "Characteristics of Modern Manufacturing Techniques," in AM ST 05 Advanced Manufacturing Systems and Technology, CISM International Centre for Mechanical Sciences, Volume 486, 5 (2005).

[7.] J. Giboz, T. Copponnex, and P. Mele, J. Micromech Microeng., 17, R96 (2007).

[8.] J.W. Mann, Plast. Eng., 30, 25 (1974).

[9.] P.D. Coates and R.G. Speight, Proc. Inst. Meclt. Eng., Part B: J. Eng. Manuf, 209, 357 (1995).

[10.] R. Dubay, A.C. Bell, and Y.P. Gupta, Polym. Eng. Sci., 37, 1550 (1997).

[11.] R.I. Tanner, Engineering Rheology, Oxford University Press (2000).

[12.] A.Y. Malkin and A.I. Isayev, Rheology: Concepts, Methods, and Applications, ChemTec Publishing, Toronto (2005).

[13.] C. Maier, Polym. Eng. Sci., 36, 1502 (1996).

[14.] G. Gordon, D.O. Kazmer, R.X. Gao, Z. Fan, and N.A. Zanjani, Temperature Modeling for a Multivariate Injection Molding Sensor, Presented at the Society of Plastics Engineers (SPE) ANTEC, Cincinnati, OH (2013).

[15.] D. Kazmer and P. Barkan, Polym. Eng. Sci., 37, 1865 (1997).

[16.] A. Peterlin, J. Polym. Sci., 8, 621 (1952).

[17.] X. Chen, L. Zhang, X. Kong, J. Lu, and F. Gao, Polym. Eng. Sci., 50, 1358 (2010).

[18.] J. Bicerano, Prediction of Polymer Properties, CRC Press, Florida 2002.

[19.] Y. Wang, Y. Xiao, Q. Zhang, X.L. Gao, and Q. Fu, Polymer, 44, 1469 (2003).

[20.] M.K. Fisches, D.O. Kazmer, and G.W. Gordon, "Verification of Rheological Mixing Rules to the Application of Masterbatches," Presented at the Society of Plastics Engineers (SPE) ANTEC, Cincinnati, OH (2013).

Guthrie Gordon, (1) David O. Kazmer, (1) Xinyao Tang, (2) Zhoayan Fan, (2) Robert X. Gao (2)

(1) Department of Plastics Engineering, University of Massachusetts Lowell, Lowell, Massachusetts 01854

(2) Department of Mechanical Engineering, University of Connecticut, Storrs, Connecticut 06269

Correspondence to: G. Gordon: e-mail: Guthrie_gordon@student.uml.edu Contract grant sponsor: National Science Foundation; contract grant number: CMMI-1000816/1000551.

DOI10.1002/pen.24170

Published online in Wiley Online Library (wileyonlinelibrary.com).

TABLE 1. Material compounding.

Material               Additive                    Color

Dow Styron 478--HIPS   --                          White (neat)
                       ECM Plastics--CPS 910       Blue
                       ECM Plastics--CHIPS 1137    Black
                       ECM Plastics--CHIPS 1150    Purple

TABLE 2. Extruder temperature profile.

Zone     Temperature ([degrees]C)

1                  180
2                  185
3                  190
4                  195
Die                200

TABLE 3. 12 run DOE, 10 samples for each run.

Run   Additive     Color    [V.sub.inject] (mm/s)

1         --       Neat              100
2         --       Neat              400
3         --       Neat             1000
4     CPS 910      Blue              100
5     CPS 910      Blue              400
6     CPS 910      Blue             1000
7     CHIPS 1137   Black             100
8     CHIPS 1137   Black             400
9     CHIPS 1137   Black            1000
10    CHIPS 1150   Purple            100
11    CHIPS 1150   Purple            400
12    CHIPS 1150   Purple           1000
COPYRIGHT 2015 Society of Plastics Engineers, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Gordon, Guthrie; Kazmer, David O.; Tang, Xinyao; Fan, Zhoayan; Gao, Robert X.
Publication:Polymer Engineering and Science
Article Type:Report
Geographic Code:1USA
Date:Dec 1, 2015
Words:4410
Previous Article:Enhanced mechanical properties, thermal stability of phenolic-formaldehyde foam/silica nanocomposites via in situ polymerization.
Next Article:Effect of mixing poly(lactic acid) With glycidyl methacrylate grafted poly(ethylene octene) on optical and mechanical properties of the blown films.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters