This action might not be possible to undo. Are you sure you want to continue?
By: Jack Peters
Understanding the basics and fundamentals of vibration analysis are very important in forming a solid background to analyze problems on rotating machinery. Switching between time and frequency is a common tool used for analysis. Because the frequency spectrum is derived from the data in the time domain, the relationship between time and frequency is very important. Units of acceleration, velocity, and displacement are typical. Additional terms such as peak-peak, peak, and rms. are often used. Switching units correctly, and keeping terms straight is a must. As much as possible, this training will follow the criteria as established by the Vibration Institute.
Jack D. Peters
Mass & Stiffness
Mass & Stiffness
All machines can be broken down into two specific categories. Mass & Stiffness Mass is represented by an object that wants to move or rotate. Stiffness is represented by springs or constraints of that movement.
Jack D. Peters
Mass & Stiffness Mass & Stiffness fn = 1/2П k/m Where: fn = natural frequency (Hz) k = stiffness (lb/in) m = mass mass = weight/gravity weight (lb) gravity (386. Peters 4 .1 in/sec2) Jack D.
Mass & Stiffness Concept ! fn = 1/2П k/m If k increases Then f increases If k decreases Then f decreases Jack D. Peters 5 .
Peters 6 .Mass & Stiffness Concept ! fn = 1/2П k/m If m increases Then f decreases If m decreases Then f increases Jack D.
Spectrum What’s This ? 1 0.0002 inch Peak Magnitude 0 0 Hz 100 Hz Jack D. Peters 7 .
Spectrum FFT.0002 inch Peak Magnitude 0 0 Hz 100 Hz Jack D. Power Spectrum 1 0. Frequency Spectrum. Peters 8 .
0002 inch Peak Y Magnitude 0 0 Hz 100 Hz X Jack D. Peters 9 .Spectrum Scaling X & Y 1 0.
Peters 1 0.0002 inch Peak Magnitude 0 0 Hz FREQUENCY 10 100 Hz .Spectrum Scaling X & Y A M P L I T U D E Jack D.
Peters .0002 inch Peak Magnitude 0 0 Hz What is it 11 100 Hz Jack D.Spectrum Scaling X & Y 1 H o w B a d I s I t 0.
Peters 12 .996094 s Jack D.Time Waveform What’s That ? 1 0.0004 inch Real -0.0004 0 s 7.
0004 0 s 7. Peters 13 .Time Waveform Time Waveform 1 0.996094 s Jack D.0004 inch Real -0.
0004 0 s 7.Time Waveform Scaling X & Y 1 0.0004 inch Y Real -0. Peters 14 .996094 s X Jack D.
996094 s 15 . Peters 1 0.0004 inch Real -0.Time Waveform Scaling X & Y A M P L I T U D E Jack D.0004 0 s TIME 7.
996094 s Jack D.0004 inch Real What is it 7. Peters 16 .Time Waveform Scaling X & Y H o w B a d I s I t -0.0004 0 s 1 0.
Time Waveform Scaling X & Y 1 H o w B a d I s I t 0. Peters .0004 0 s What is it 17 1 s Jack D.0004 inch Real -0.
996094 s 100 Hz Jack D.0004 0 s 7.0002 inch Peak Magnitude 0 0 Hz 1 0.Time & Frequency Double Trouble 1 0. Peters 18 .0004 inch Real -0.
The X Scale The X Scale Jack D. Peters 19 .
The X Scale Hertz (Hz) One Hertz (Hz) is equal to 1 cycle / second It is the most common term used in vibration analysis to describe the frequency of a disturbance. 60 cpm = 1 cps = 1 Hz Jack D. Peters 20 . Never forget the 1 cycle / second relationship ! Traditional vibration analysis quite often expresses frequency in terms of cycle / minute (cpm). This is because many pieces of process equipment have running speeds related to revolutions / minute (rpm).
The X Scale Relationship with Time The frequency domain is an expression of amplitude and individual frequencies. Jack D. Peters 21 . T(s) = 1 / F(Hz) Keep in mind that the time domain is an expression of amplitude and multiple frequencies. A single frequency can be related to time. F(Hz) = 1 / T(s) The inverse of this is also true for a single frequency.
The X Scale Concept ! If: Then: F = 1/T and T = 1/F FT = 1 Jack D. Peters 22 .
The X Scale Concept ! FT = 1 If: F increases Then: T decreases If: T increases Then: F decreases Jack D. Peters 23 .
449082 mV 100 Hz Y:706.The X Scale Single Frequency X:55 Hz Pwr Spec 1 1 V rms Magnitude 0 0 Hz X:27.46948 ms Y:3.00806 ms dX:18.8129 mV Jack D. Peters 24 .18848 ms Time 1 1 V Real -1 0 s 62.579427 mV dY:2.
7825 mV 100 Hz Y:706.9266 mV 100 Hz Y:706.8129 mV Jack D.The X Scale Multiple Frequencies X:55 Hz Pwr Spec 1 1 0 Hz X:78 Hz Pwr Spec 1 1 0 Hz X:21 Hz Pwr Spec 1 1 0 Hz X:42 Hz Pwr Spec 1 1 0 Hz 100 Hz Y:706. Peters 25 .9236 mV 100 Hz Y:706.
46948 ms 62.46948 ms Jack D.The X Scale Multiple Time Time 55 1 1 V 0 s Time 78 1 1 V 0 s Time 21 1 1 V 0 s Time 42 1 1 V 0 s 62.46948 ms 62.46948 ms 62. Peters 26 .
46948 ms Jack D. Peters 27 .The X Scale Real Life Time 55 + 78 + 21 + 42 = Trouble ! TIME 1 4 V Real -4 0 s 62.
Peters 28 .46948 ms Jack D.8129 mV Y:706.7825 mV Y:706.9236 mV 62.9266 mV Y:706.The X Scale Frequency Spectrum TIME 1 4 V Real -4 0 s X:21 Hz X:42 Hz X:55 Hz X:78 Hz FREQUENCY 1 1 V rms 0 Hz 100 Hz Y:706.
Peters 29 .The X Scale The Most Copied Slide in the History of Vibration Analysis ! Amplitude Inp ut Tim e Tim eW Fr cy uen eq um ectr Sp ave for m Jack D.
0 0 Hz 100 Hz Jack D. 800. 400. or the X scale is broken down into 800 points. 200. Peters 30 . and 3200 lines are common choices.0002 inch Peak Magnitude This spectrum has 800 lines. 1600. 100.The X Scale Lines of Resolution The FFT always has a defined number of lines of resolution. 1 0.
The X Scale Filter Windows • Window filters are applied to the time waveform data to simulate data that starts and stops at zero. • We still like window filters ! Jack D. • They will cause errors in the time waveform and frequency spectrum. Peters 31 .
Peters 32 .The X Scale Window Comparisons Jack D.
The X Scale Filter Windows • • • • Hanning (Frequency) Flat Top (Amplitude) Uniform (No Window) Force Exponential (Frequency Response) Hanning 16% Amplitude Error • Force/Expo Set-up Window functions courtesy of Agilent “The Fundamentals of Signal Analysis” Application Note #AN 243 Flat Top 1% Amplitude Error Jack D. Peters 33 .
Peters 34 .The X Scale Filter Windows • Use the Hanning Window for normal vibration monitoring (Frequency) • Use the Flat Top Window for calibration and accuracy (Amplitude) • Use the Uniform Window for bump testing and resonance checks (No Window) Jack D.
The number of analyzer lines depends on the analyzer and how the operator has set it up. Example: 0 . Peters 35 .The X Scale Minimum Derived Hz The minimum derived frequency is determined by: Frequency Span / Number of Analyzer Lines (data points) The frequency span is calculated as the ending frequency minus the starting frequency.5 Hz / Line Jack D.0) / 800 = 0.400 Hz using 800 lines Answer = (400 .
5 = 0.0 Hanning Window Function = 1. Peters 36 Note: More discussion later on window functions for the analyzer ! .400 Hz using 800 Lines & Hanning Window Answer = (400 / 800) 1.75 Hz / Line Jack D.5 Flat Top Window Function = 3.The X Scale Bandwidth The Bandwidth can be defined by: (Frequency Span / Analyzer Lines) Window Function Uniform Window Function = 1.5 Example: 0 .
Peters 37 .5 Hz / Line Jack D.The X Scale Resolution The frequency resolution is defined in the following manner: 2 (Frequency Span / Analyzer Lines) Window Function or Resolution = 2 (Bandwidth) Example: 0 .400 Hz using 800 Lines & Hanning Window Answer = 2 (400 / 800) 1.5 = 1.
The instructor suggests a hanning window and 800 lines. Peters 38 . What frequency span is required to accurately measure these two frequency disturbances ? Jack D. Frequency #1 = 29.5 Hz. Frequency #2 = 30 Hz.The X Scale Using Resolution The student wishes to measure two frequency disturbances that are very close together.
5 = 2 (Frequency Span / 800) 1.29. Peters 39 .5 0.5 Hz / Line Resolution = 2 (Bandwidth) BW = (Frequency Span / Analyzer Lines) Window Function Resolution = 2 (Frequency Span / 800) 1.The X Scale Using Resolution Resolution = 30 .5 = 0.5 = 3 (Frequency Span) / 800 400 = 3 (Frequency Span) 133 Hz = Frequency Span Jack D.5 0.
5 seconds Jack D.The X Scale Data Sampling Time Data sampling time is the amount of time required to take one record or sample of data. It is dependent on the frequency span and the number of analyzer lines being used. Peters 40 . TSample = Nlines / Fspan Using 400 lines with a 800 Hz frequency span will require: 400 / 800 = 0.
The X Scale Average & Overlap • Average .50% TR#1 TR#2 TR#3 FFT#1 FFT#2 FFT#3 0% Overlap TR#1 How long will it take for 10 averages at 75% overlap using a 800 line analyzer and a 200 Hz frequency span? TR#2 TR#3 FFT#1 50% Overlap FFT#2 FFT#3 Jack D.On • Overlap Percent . Peters 41 .
The X Scale
75% Overlap ?
• • • • 10 Averages 75% Overlap 800 Lines 200 Hz
Average #1 = 800 / 200 Average #1 = 4 seconds Average #2 - #10 = (4 x 0.25) Average #2 - #10 = 1 second each Total time = 4 + (1 x 9) Total time = 13 seconds
Jack D. Peters 42
The Y Scale
The Y Scale
Jack D. Peters
The Y Scale
The “Y” scale provides the amplitude value for each signal or frequency. Default units for the “Y” scale are volts RMS. Volts is an Engineering Unit (EU). RMS is one of three suffixes meant to confuse you ! The other two are: (Peak) and (Peak - Peak)
Jack D. Peters 44
46948 ms Y:-993.Peak value is expressed from the peak to peak amplitude.8563 mV dY:1. The spectrum value uses the suffix “Pk-Pk” to denote this.999169 V Jack D. Peters 45 .994871 V 100 Hz Y:1.43042 ms dX:9.Peak) The Peak . X:55 Hz Pwr Spec 1 2 V Pk-Pk Magnitude 0 0 Hz X:22.The Y Scale Pk-Pk (Peak .094238 ms Time 1 1 V Real -1 0 s 62.
516602 ms Time 1 1 V Real -1 0 s 62. The spectrum value uses the suffix “Peak” to denote this.The Y Scale Pk (Peak) The time wave has not changed.46948 ms Y:3.5843 mV Jack D. X:55 Hz Pwr Spec 1 1 V Peak Magnitude 0 0 Hz X:27.4356 mV 100 Hz Y:999. The Peak value is expressed from zero to the peak amplitude. Peters 46 .00806 ms dX:4.579427 mV dY:997.
Jack D.The Y Scale RMS (Root Mean Square) The time wave has not changed.00806 ms dX:2.8129 mV 100 Hz Y:3. The spectrum value uses the suffix “RMS” to denote this.579427 mV dY:709.1976 mV 62. Peters X:55 Hz Pwr Spec 1 1 V rms Magnitude 0 0 Hz X:27.288818 ms Time 1 1 V Real -1 0 s Y:706.46948 ms 47 . The RMS value is expressed from zero to 70.7% of the peak amplitude.
288818 ms Time 1 Y:3.00806 ms dX:2.094238 ms Time 1 62.4356 m 2 V Peak Magnitude 0 0 s X:22.579427 mV dY:709.5843 mV X:55 Hz Pwr Spec 1 Y:706.00806 ms dX:4.The Y Scale Suffix Comparison X:27.46948 ms 0 Hz 100 Hz 0 Hz X:55 Hz Pwr Spec 1 100 Hz Y:1.8129 mV RMS 1 V Real -1 Peak 1 V Real -1 Peak . Peters 48 .994871 V 2 V Pk-Pk Magnitude 0 0 s 62.46948 ms Y:-993.516602 ms Time 1 62.999169 V 0 Hz X:55 Hz Pwr Spec 1 100 Hz Y:999.43042 ms dX:9.46948 ms Y:3.1976 m 2 V rms Magnitude 0 0 s X:27.8563 mV dY:1.Peak 1 V Real -1 Jack D.579427 mV dY:997.
707 = RMS RMS x 1.The Y Scale Changing Suffixes Many times it is necessary to change between suffixes. Peters 49 .414 = Peak Peak x 2 = Pk-Pk Jack D. Pk-Pk / 2 = Peak Peak x 0.
Acceleration = g’s Peak or rms. Note: 1 mil = 0.Peak Velocity = in/s Peak or rms. Peters 50 .001 inches Jack D.. what is the standard ? Vibration Institute: Displacement = mils Peak ..The Y Scale Standard Suffixes Now that we have learned all about the three standard suffixes that might possibly confuse the “Y” scale values.
Peters 20 mV / Pa 200 mV / mil 10 mV / fpm 10 mV / V 51 . it is possible to incorporate a unit proportional to volts that will have greater meaning to the user. Examples: 100 mV / g 1 V / in/s 50 mV / psi 33 mV / % Jack D. Instead of the default “volts”.The Y Scale Engineering Units (EU) Engineering units are used to give meaning to the amplitude of the measurement.
The Y Scale EU’s the Hard Way Sometimes we forget to use EU’s. or just don’t understand how to set up the analyzer. Peters 52 . There is no immediate need to panic if ???? You know what the EU is for the sensor you are using. Example: An accelerometer outputs 100 mV / g and there is a 10 mV peak in the frequency spectrum. What is the amplitude in g’s ? Answer = 10 mV / 100 mV = 0.1 g Jack D.
The Y Scale The Big Three EU’s Acceleration Velocity Displacement Jack D. Peters 53 .
but the model calls for displacement.The Y Scale Converting the Big 3 In many cases we are confronted with Acceleration. or Displacement. How do we change between these EU’s ? Jack D. Peters 54 . Velocity. but the manufacturer quoted the equipment specifications in velocity. Maybe we have taken the measurement in acceleration. Maybe we have taken the data in displacement. but are not happy with it.
2 feet/second2 32. Peters 55 .The Y Scale 386.1 What ? 1g = 32.2 feet 2 second X 12 inches foot 2 386.1 inches/second g Jack D.
1 x 2(Pi)f Velocity (inch/s) x 2(Pi)f Displacement (inch) Jack D.1 Acceleration (g’s) x 386. Peters 56 ÷ 2(Pi)f ÷ 2(Pi)f .The Y Scale Go With The Flow I ÷ 386.
The Y Scale Metric Go With The Flow I ÷ 9807 Acceleration (g’s) x 9807 x 2(Pi)f Velocity (mm/s) x 2(Pi)f Displacement (mm) Jack D. Peters 57 ÷ 2(Pi)f ÷ 2(Pi)f .
Peters 58 ÷2 x .Peak x2 Peak x 1.414 RMS Jack D.707 .The Y Scale Go With The Flow II Peak .
What is the velocity ? 2π x 25 cycles second 0. Peters 59 .5g x 386.1 inches 2π x 25 cycles second cycle 1.The Y Scale Doing the Math Units 0.1 inches second2 g There is a 0.5g x 386.23 inches/second Jack D.5 g vibration at 25 Hz.5 x 386.1 inches g second2 x 1 second 2π x 25 cycles 0.
017 x 1.1) / (6.Velocity Example: Find the equivalent peak velocity for a 25 Hz vibration at 7 mg RMS ? = (g x 386. Peters 60 .007 x 386.414 = 0.024 inches / second Pk Jack D.28 x 25) = 0.017 inches / second RMS Answer = 0.The Y Scale Acceleration .1) / (2 Pi x F) = (0.
Peters 61 .000153 x 2 = 0.024 / (6.The Y Scale Velocity .000153 inches Pk Answer = 0.000306 inches Pk-Pk Jack D.28 x 25) = 0.024 in/s Pk ? = Velocity / (2 Pi x F) = 0.Displacement Example: Find the equivalent pk-pk displacement for a 25 Hz vibration at 0.
414) 2 = 0.28 x 52)2 = 0. Peters 62 .1) / (2 Pi x F)2 = (0.000154 inches Pk-Pk Jack D.015 x 386.000054 x 1.1) / (6.Displacement Example: Find the equivalent Pk-Pk displacement for a 52 Hz vibration at 15 mg RMS ? = (g x 386.The Y Scale Acceleration .000054 inches RMS Answer = (0.
Sensors Sensors Jack D. Peters 63 .
Peters 64 .Sensors Accelerometers • Integrated Circuit Electronics inside Industrial • Charge Mode Charge Amplifier Test & Measurement Jack D.
Peters 65 .Sensors Accelerometer Advantages • • • • • • Measures casing vibration Measures absolute motion Can integrate to Velocity output Easy to mount Large range of frequency response Available in many configurations Jack D.
Sensors Accelerometer Disadvantages • Does not measure shaft vibration • Sensitive to mounting techniques and surface conditions • Difficult to perform calibration check • Double integration to displacement often causes low frequency noise • One accelerometer does not fit all applications Jack D. Peters 66 .
Peters 67 .Sensors Mass & Charge Relative movement between post & mass creates shear in ceramic producing charge. Mass Ceramic/Quartz Post Jack D.
Peters 68 .Sensors Accelerometer Parameters Performance Suited for Application Sensitivity (mV/g) Frequency Response of target (f span) Dynamic Range of target (g level) Jack D.
000 15. Peters 69 .500 10.Sensors Mounting the Accelerometer 500 2000 6. Jack D.000 15.000 Hz.
Peters 70 .000 Frequency. mounting might not be as good as the manufacturer had in the lab ! What happened to the high frequency ? 100 1. Hz 10.000 Jack D.Sensors Realistic Mounting Stud Bees Wax Adhesive Magnet Hand Held In the real world.
Sensors Mounting Location Vertical Load Zone Radial Vertical Horizontal Axial Axial Horizontal Jack D. Peters 71 .
Peters 72 .Sensors Mounting Location Load Zone Radial Vertical Horizontal Axial Jack D.
28 > 0.24 .7 Acceptance of new or repaired equipment Unrestricted operation (normal) Surveillance Unsuitable for Operation Note #1: The rms velocity (in/sec) is the band power or band energy calculated in the frequency spectrum.7 > 0.Sensors Accelerometer Alarms Machine Condition Velocity Limit rms peak < 0. Peters 73 .24 0. Note #2: The peak velocity (in/sec) is the largest positive or negative peak measured in the time waveform. Jack D.12 .0.28 < 0.0.08 < 0.12 0.16 < 0.
Peters 74 .Sensors Proximity Probes Jack D.
the output voltage of the driver also varies. Target Jack D. A small dc voltage indicates that the target is close to the probe tip. As the power varies with target movement in the radio frequency field. Peters The variation of dc voltage is the dynamic signal indicating the vibration or displacement. 75 . eddy currents are generated on the surface of the target. A large dc voltage indicates that the target is far away from the probe tip. and power is drained from the radio frequency signal.Sensors Proximity Probe Theory Driver Cable Probe The tip of the probe broadcasts a radio frequency signal into the surrounding area as a magnetic field. If a conductive target intercepts the magnetic field.
4340 380 mV/mil 370 mV/mil 330 mV/mil 290 mV/mil 250 mV/mil 200 mV/mil • Depends on probe. and driver.Sensors Output Values • Typical – 100 mv/mil – 200 mv/mil • • • • • • Calibration Examples Copper Aluminum Brass Tungsten Carbide Stainless Steel Steel 4140. Jack D. Peters Based on typical output sensitivity of 200 mV/mil. 76 . • Target material varies output. cable (length).
Peters 77 .Sensors Proximity Probes .Advantages • • • • • • Non-contact Measure shaft dynamic motion Measure shaft static position (gap) Flat frequency response dc – 1KHz Simple calibration Suitable for harsh environments Jack D.
cable & driver) • Must have relief at sensing tip from surrounding metal Jack D.Disadvantages • • • • • Probe can move (vibrate) Doesn’t work on all metals Plated shafts may give false measurement Measures nicks & tool marks in shaft Must be replaced as a unit (probe.Sensors Proximity Probes . Peters 78 .
Peters 79 .Sensors Typical Mounting vertical probe horizontal probe H V time Facing Driver to Driven (independent of rotation) Jack D.
Peters 80 .Sensors Looking at Orbits Vertical for Amplitude Horizontal for Time Base Jack D.
Sensors The Orbit Display Mils ils Mils Jack D. Peters 81 .
Sensors Unbalance Jack D. Peters 82 .
Peters 83 .Sensors Unbalance with Orbit Jack D.
Sensors Misalignment Jack D. Peters 84 .
Peters 85 .Sensors Misalignment with Orbit Jack D.
Sensors And the problem is ? Jack D. Peters 86 .
5 0.000 RPM 0.5 0. Jack D.3 .3 0.2 0.600 RPM < 10.6 Normal Surveillance Planned Shutdown Unsuitable for Operation Note #1: R is the relative displacement of the shaft measured by either probe in mils peak-peak. Note #2: C is the diametrical clearance (difference between shaft OD and journal ID) measured in mils.Sensors Proximity Probe Alarms Machine Condition Allowable R/C < 3. Peters 87 .2 .0.0.4 0.7 0.4 0.
Analyzer Input - Front End
• Coupling - AC, DC
AC coupling will block the DC voltage. It creates an amplitude error below 1 Hz. DC coupling has no error below 1 Hz, but the analyzer must range on the total signal amplitude.
Analyzer Input Capacitor Blocks DC
AC Coupling Capacitor in DSA
• Antialias Filter - On, Off
Prevents frequencies that are greater than span from wrapping around in the spectrum.
If the antialias filter is turned off, at what frequency will 175 Hz. appear using a 0 100 Hz span, and 800 lines ? 1024/800 = 1.28 100 x 1.28 = 128 Hz 175 Hz - 128 Hz = 47 Hz. 128 Hz - 47 Hz = 81 Hz
Jack D. Peters
Low End Frequency Response
To the right is a typical problem with frequency response at the low end of the frequency spectrum. This low end roll off was a result of AC coupling on CH #2 of the analyzer. Values below 2.8 Hz are in error, and values less than 0.5 Hz should not be used.
Jack D. Peters
Input 1 0.01 V rms Magnitude 0.002 0 Hz Output 2 0.01 V rms Magnitude 0 0 Hz X:500 mHz X:2.8125 Hz Freq Resp 2:1 2 Magnitude 0 0 Hz X:5.34375 Hz Coherence 2:1 1.1 Magnitude 0.9 0 Hz 25 Hz Y:1.000001 25 Hz Y:720.0918 m Y:984.7542 m 25 Hz 25 Hz
Jack D. Peters
Peters 91 . For the Measurement that can’t get away ! Jack D.Data Collection Tape Recorders “Insurance Policy” Multi-channel digital audio tape recorders.
Jack D. Smaller portable units with 2 – 4 channel inputs and firmware operating systems. Peters 92 .Data Collection Dynamic Signal Analyzers “Test & Measurement” Large PC driven solutions with multiple channels and windows based software.
Peters 93 .Data Collection Data Collectors “Rotating Equipment” Jack D.
Peters • • • • • • Data Analysis History Trending Download to PC Alarms “Smart” algorithms 94 .Data Collection Data Collectors “Rotating Equipment” • Route Based • Frequency Spectrum • Time Waveform • Orbits • Balancing • Alignment Jack D.
Thomas. Vibration Analysis Design. Selection. and Installation. Application Note. Mounting. Robert Jr. Application note 243-1 • • Jack D. Connection Technology Center Agilent Technologies. ISBN 09669500-0-3 LaRocque. Application note 243 Agilent Technologies. Ronald L. Effective Machinery Measurements using Dynamic Signal Analyzers.Beginning Vibration Bibliography • • • Eisenmann. Machinery Malfunction Diagnosis and Correction. & Eisenmann. Robert Sr. Peters 95 .. Basic Machinery Vibrations.. ISBN 0-13-240946-1 Eshleman. The Fundamentals of Signal Analysis.
Peters 96 .ctconline.com in the “Technical Resources” section Jack D.Beginning Vibration Thank You ! You can find technical papers on this and other subjects at www.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.