Website update in progress! You might be logged out of your account. If this occurs, please log back in.

Website update in progress! You might be logged out of your account. If this occurs, please log back in.

Sign Up

Please use your business email address if applicable

Part 2 Machinery Health Monitoring Depends on Accelerometers: Signals from Accelerometers


Part 1 told you that accelerometers convert motion into electrical signals. Part 2 warns you against certain electrical errors.


Figure 1 suggests some readout instruments. Identify the domain (time or frequency) in which each functions or displays:


Figure 1: Sensors in a variety of readout instruments

The first three? The time domain. The last? The frequency domain. The computer? In both domains. Let’s focus on the computer and spectrum analyzer.

Never introduce a millivolt-level sensor electrical signal directly into electrically noisy environments of desktop/laptop computers. Instead, route it through an electrically quiet front-end amplifier.


Figure 2: Data acquisition front end


In Parts 4 and 5 of this series, we are going to look for changes in the signatures or the spectra of machines. Spectra?

Mentally hold the mechanical spectrum analyzer shown in Figures 3 and 4 against a 1750 rpm motor. Each reed was factory tuned to resonate at a specific frequency. One, tuned to 1750 rpm, is responding strongly.


Figure 3: Handheld mechanical spectrum analyzer contacts motor turning at 1750 rpm

Figure 4 suggests the internal construction.


Figure 4: Internal construction of reed device

Hold the analyzer quietly, all reeds at zero. Knuckle strike the analyzer bottom. If Figure 5 were a video, you’d see all the reeds responding simultaneously, each at its own frequency.


Figure 5: Multi-reed tachometer just struck

Agree? The shock event contains energy at many frequencies. If each reed could record how vigorously it responded, we’d have a spectrum of that event. (It’s called the shock response spectrum (SRS.) We’re preparing to use machinery spectra in Parts 4 and 5 of this series.


Let’s move on. In Parts 4 and 5, we’ll determine how much of our machine’s overall vibration (accelerometer-sensed) occurs at various frequencies. We interpose numerous fixed electrical filters between our accelerometer and our readout instrument.

Back in the 1920s, the first fixed audio filters were an octave wide. Example: 37.5 to 75 Hz, showing little detail (see Figures 6 and 7).

Figure 6: Octave band analyzer


Figure 7: Set of octave band filters

One-third octave analyzers (see Figures 8 and 9) suffice for noise code enforcement. As machinery speeds and vibration frequencies have increased, early analyzers have become technically obsolete, although many are still in use.

One-third octave filters plus microphone, amplification, weighting networks and readout meter are combined in Figure 9 into a sound level meter (SLM).


Figure 8: Set of 1/3 octave filters


Figure 9: Sound level meter


How can we accurately determine the magnitudes and frequencies of the several vibrations our machine generates simultaneously? Sometimes, these are very close together in frequency.

Figure 10’s historic technician manually tunes his analog analyzer across a pump’s range of vibration frequencies. When he identifies a spectral peak, he writes down its height and frequency. Very time-consuming.


Figure 10: Example of analog analysis


Figure 11: Example of a dedicated spectrum analyzer (Courtesy of Ono Sokki)

Today’s digital sound and vibration spectrum analyzers are much faster. Perhaps your vibration band of interest is 0 to 500 Hz. If your analyzer provides 500 windows or 500 lines, your resolution is 1 Hz. That is, you can measure and plot what vibration is occurring in each of the windows, 398-399, 399-400, 400-401 Hz., etc.


Reasonable accuracy is important. But errors can and do creep in.


Figure 12: Engine test cell suggests some sources of electrical noise

Most measurement and analysis of vibration signals are done digitally using computers. Analog signals from accelerometers, force sensors, etc., must be digitized or sampled. If we are not extremely careful, we can get wrong data. We will consider aliasing and leakage in Part 4.

But plenty of other error sources exist. Electrical error sources mentioned here affect both analog and digital analysis.

Avoid noise or unwanted signals that can contaminate your measurements. Figure 13 suggests the idea that noise or unwanted signals make information gathering difficult.


Figure 13: Electrical noise makes information hard to read


The first piezoelectric (PE) accelerometers (circa 1955) fed signals into vacuum tube (later solid state) voltage amplifiers. Most of today’s PE accelerometer signal conditioners are called charge amplifiers or, more accurately, charge to voltage converters.

Signal conditioners used with pressure (PR) and specific gravity (SG) sensors must also supply DC excitation to the resistive bridge elements in the sensors. Figure 14 suggests the functions needed for each channel of instrumentation.


Figure 14: One channel of instrumentation

In-line charge amplifiers (Figures 15 and 16) have given way to today’s inside the accelerometer amplification.


Figure 15: Charge amplifiers (Courtesy of Element & Steve Brenner)


Figure 16: In-line charge amplifier


The concept of troublesome ground loops is illustrated in Figure 17. A PE accelerometer feeds a signal conditioner and readout device, here called measuring instruments. They are connected to local ground at different locations on an electric power distribution system or water piping. Unfortunately, those two grounds are some millivolts, or volts different in potential, because of unknown circulating currents in various conductors.


Figure 17: Ground loops

If your situation resembles Figure 17, unground either your sensor or your measuring instrument.

Never allow your cables to become kinked, twisted, or knotted. Never stress them or flop them loosely around. Don’t step on or drive over them. Don’t contaminate them with chemicals, dirt, oil, or grease.

If you suspect a cable is bad, destroy it and obtain a new replacement. Your data is so expensive that you can’t risk its validity by using suspect cables.

Don’t leave your entire cable (and accelerometer) inventory available to just anyone. Keep new equipment hidden until needed.

Correct cable routing lessens electrical interference. Don’t run cables close to AC power lines. If crossing a power line, do so at 90 degrees.

Any high electrical impedance signal lead from sensor to signal conditioner is vulnerable to electrical noise, so keep it short. Put the signal conditioner close to the sensor. It’s okay to have a long lead after the signal conditioner because that lead is at low electrical impedance, perhaps 100 ohms.


Carrying Figure 18 to a long-awaited conclusion, manufacturers have been placing microcircuit amplifiers inside accelerometer housings for some years (see Figure 19).


Figure 18: Keep high Z wiring short


Figure 19: Amplification inside sensor case

High impedance circuitry is sealed inside the sensor case, untroubled by contaminated environments and by electrical interference. It can drive long cables without signal degradation or loss of resolution. It works directly into modern fast Fourier transforms (FFTs) and data collectors.


Unfortunately, most such sensors can only operate at -65 to +250 degrees F.


In the first article of this series, we looked at several accelerometer types that currently are popular. Our concern was mostly on mechanical aspects. In this article, our emphasis is on the associated electronics. In the third article, we will consider calibrating these sensors and systems.

Keep reading...Show less
ChatGPT with
Find Your Answers Fast