Understanding & Analyzing Signals

Understanding Signals

• In-phase waves add up point by point, resulting in a sine wave with the same frequency but increased amplitude
• Waves 180 degrees out of phase add up to zero, acting as mirror images, as used in rotor balancing and noise-canceling headphones

Beating of Signals

• Signals starting in phase gradually drift out of phase and back, causing beating
• In-phase amplitudes add, while out-of-phase amplitudes partially cancel out
• Beat frequency calculated by the difference between the two frequencies
• Beat Period 1/beat frequency

Time Waveform Analysis

• Period, marked by horizontal arrow, represents 0.01 seconds, indicating a frequency of 100 Hz
• Expect a peak at 100 Hz and harmonics if the wave is repetitive but not a pure sine wave
• Higher frequency can be found from Period = 0.01s using Frequency (Hz) = 1 / Period (s) equation, and the spectrum will have a peak at 100 Hz

Modulating Frequency

• Modulating frequency period calculation: period equalling 20 x 0.01s = 0.2 seconds and frequency equalling 1/0.2 = 5 Hz
• Expect a peak at 100 Hz with sidebands at 95 and 105 Hz

Noise Floor in Spectrum

• The primary causes include impacts or single events and random vibration
• Random vibration originations include loose sensors, parts, flow noise, or turbulence
• Impact or single event common causes: Cavitation, late stage bearing wear
• Analyze the time waveform

Amplitude Modulation (AM)

• AM results in sidebands in the spectrum, often associated with gears, bearings, and AC motors
• Confusing AM in the time waveform can be distinguished from beating
• Beating leads to two closely spaced peaks in the vibration spectrum
• Beating caused by adding two sine waves with slightly different frequencies
• AM involves modulating a sine wave by another sine wave

Signal Processing-Low Pass Filter

• Low pass filter permits frequencies under the filter frequency and reduces, but does not eliminate, those in the transition band to prevent aliasing
• A stop band describes totally eliminated / filtered out data
• “fc” the half power point, has a of 3dB amplitude reduction
• frequencies below “fc” the pass band
• ISO 10816-3 specifies RMS overall level measurement between 10 Hz and 1000 Hz for machines exceeding 600 RPM, which implies use of band pass filter
• A high pass filter removes amplified low frequencies during integration
• A low pass filter manages the Fmax setting

Integration

• Integration converts acceleration to velocity or velocity to displacement
• Integration can be performed on the analog or digital signal after vibration sampling
• Integration attenuates levels at frequencies over the velocity and acceleration lines
• Velocity amplification by 1/f² at low frequencies requires a high pass filter to remove noise
• Analog integration can enhance signal-to-noise ratio if dynamic range is limited

Digital Integration

• Modern 24-bit data collectors perform digital integration, eliminating the need to amplify or attenuate signals
• Analog integration results in velocity data before sampling, while digital integration yields acceleration data
• Is useful for time waveform acceleration data

Signal Digitization

• Adjusting before digitizing aligns voltage with the A/D converter, maximizing pixel resolution
• Analog integration adjusts voltage
• Auto ranging adjusts the input signal
• Auto ranging is functional for stable signals
• Turning auto ranging off gives a more accurate reading if used with a bump sensor test

Setting Fmax

• Setting the Fmax is 2.56x Fmax and frequencies above Fmax are filtered out, preventing aliasing
• Earlier data collectors used analog filters, limiting data collector optiona
• The delta sigma method provides digital low pass filter is passed, then re-sampled with Fmax allowing mor options in Fmax setting

Dynamic Range

• It is the ratio of measurable signal amplitudes
• Dynamic range = 20 x log (biggest/smallest signal)
• 12 bits gives 4096 values, equalling 72 dB
• Sensors also have a dynamic range

Signal to Noise Ratio

• Describes the signals size against background noise
• Relates to signal properties
• Great data won't solve a poor measurement

Sampling

• Time 0.08 seconds at T = LOR/Fmax = 1600/2000 = 0.8 seconds, and software analysis occurs during waveform

Best Practices for Sampling with Software

• Measurement time multiplied by ten for a ten average to get 8 seconds
• 50% Overlap reuses only half of nine time blocks to get a more accurate 4.4 second total time

Four Pole Motors

• In a 60 Hz Environment they display a synchronous speed of 1800 RPM/60 = 30 Hz, with slip at 30 RPM/60 = 0.5 Hz, generating a pole pass frequency PPF = 4 poles x 0.5 = 2 Hz

Windowing factor

• For 50 Orders Fmax is equal to 50 x 30 Hz = 1500 Hz
• Hanning windows have a 1.5 factor
• Separating Frequency SF > 2 x (Fmax/LOR) x WF requires shifting LOR to the left of the equation with these steps LOR/SF > 2 x Fmax x WF or now known 3200.00
• Available Option for LOR requires reevaluation of measurements in LOR, with updated Fmax
• Final averaging time for 4 2.13 second tests to 8.53 seconds

Averaging Types for testing

• Bump test benefits from Linear averaging
• Normal route test benefits from Peak hold averaging
• Special gearbox test requires a Negative averaging
• Bump test on Machine in motion requires Exponential averaging
• In the Averaging process Give more weight to newer data
• Collect the highest vibration amplitudes on a machine when tested over a period of time

10% Pre Triggered Measurements

• Means the Hammer channel buffers reading from hammer channel
• Will provide an output to channel to ring down
• The hammer captures the hit, auto ranging off

Autocorrelation

• Autocorrelation: quantifies a signal’s similarity to its time-shifted version
• Useful for extracting hidden periodicities from waveforms
• Keeps data in time format
• Revealing data, and periodic rates

Order Tracking

• Tracks deal with speed
• If Order tracking used
• Try testing at consistent speeds, or staggering averages

Issues With Data Collection

• Changing Speed smears data
• Waterfalls improve this data's resolution

Order Tracking

• Use Order tracking as a wind turbine
• It is measured in orders with the shaft speed, and can use zebra tape
• Time synchronous averaging requires rate synchronized with Tacos
• 2x LF will disappear in the Bearings