Accelerometers
In the field Instrumentation and Test, we are often forced to use proxies. We can’t measure what we need to know, either because the technology to do it does not exist or because it would be impractical to do so. When situations like this arise, we need to find something else we can measure that gives us insight into the physical phenomena that we are trying to understand.
Very frequently this is the case when dealing with machinery, particularly engines and gearboxes. Placing any sensors within the heart of the engine, amidst the chaos of churning metal and heat, is challenging. This is even more true if the engine is being retrofitted and has not been designed with embedded instrumentation in mind.
So, if we want to monitor a piece of machinery, a pump, an engine, or a gearbox, how do we do it? Typically, mechanical objects create vibrations, and if they’re broken or close to breaking, they create more vibrations than usual. These vibrations will occur at characteristic frequencies, as gear teeth mesh, as fan blades pass, and as camshafts spin. We can use these frequencies to gain significant insight into the function of the machine.
To do this, we use accelerometers. Accelerometers come in many shapes and sizes, but the vast majority boil down to a weight on a spring. Something can only accelerate if it’s subject to a force; therefore, for the mass to accelerate, it must be placed under a force via the spring. The spring deflects linearly based on how much force it is subject to, and we therefore get the linear relationship between spring deflection and acceleration. Using Hooke’s law and the spring constant, as well as the known mass of the weight, we can establish what a given deflection corresponds to in terms of acceleration along the axis of the spring. The trick, therefore, and what differentiates most accelerometer types, is the technology used to measure that deflection.
There are three typical approaches: capacitive, piezoelectric, and piezoresistive. Capacitive accelerometers measure the change in capacitance between two plates, one mounted to the mass, one stationary, as the mass moves under acceleration. The challenge with this is measuring the capacitance change. The typical method to measure capacitance is to charge a capacitor and then discharge it, measuring the voltage characteristics of the charge-discharge signal. This takes time and may be challenging when dealing with very high-frequency vibration. As such, capacitance-based accelerometers are typically used to measure static acceleration. This is used to measure gross acceleration, such as for IMU analysis or orientation finding.
Piezo-resistive based accelerometers have a piezo-crystal mounted against the mass, such that when the mass deflects, it changes the resistance of the crystal. Resistance change can be measured using a Wheatstone bridge. Piezo-resistance change has a very high response time and therefore can be used for the measurement of dynamic acceleration, but also static. They are more sensitive to temperature effects than capacitive and piezo-electric accelerometers, but with modern techniques, temperature effects can be compensated for, at a cost.
Piezo-electric accelerometers can only measure dynamic acceleration. Under static acceleration, the mass deflects, generating a potential difference across the attached piezo crystal, but this is quickly dissipated as an equalising current flows. Therefore, if acceleration remains constant over a period of seconds, the accelerometer will return to showing zero acceleration even if it has remained constant. However, dynamic acceleration, be it vibration or shock, is a very common use case, and this can be measured very accurately as an alternating current is generated.
The key challenge with these accelerometers is frequency response. Masses on springs have resonant frequencies where the output will differ massively from the input due to resonance. To tackle this, accelerometers are typically damped and supplied with a known frequency response curve to allow it to be compensated for. Manufacturers carefully design their accelerometers to have as close to linear response curves in particular regions as possible. Most accelerometers designed for dynamic work will come with a quoted frequency range, such as 0-20 kHz, where their response is within a few percent of actual. They will still respond to frequencies outside of this range, but the response will not be within tolerance.
Returning to the idea of a proxy, what if we have a wind turbine gearbox with a 60 to 1 ratio from the main shaft that we need to understand? You might naturally think that the highest frequency we’d need to measure would be sixty times the rotation speed of the main hub, a result probably in the hundreds of Hz. This would capture the frequencies of rotation of each of the main gears and the output and input shaft. We could monitor and see how the energy in each frequency level is changing over time. However, typically we also want to gain an understanding of how the gears are meshing, which means we need to resolve the gear mesh frequency. If there are 100 teeth on one of the gears, that puts the highest frequency needing to be measured up by another hundred, now we’re in the tens of kilohertz range. With gas turbine engines, where geared turbo-fans are starting to gain popularity and you’re dealing with multiple revolutions per second in the hundreds, you can see that we might need accelerometers capable of resolving frequencies even higher. This is before we’ve even got into sample rate and the Nyquist frequency.
Flintmore has extensive expertise in working with accelerometers, both in analysis for our IMU analysis and fall detector projects and in the analysis of rotating machinery. Gaining an understanding of the internals of a gas turbine is highly challenging. By placing accelerometers at key points of the turbine and by analysing particular frequency bands, the behaviour of each shaft, the compressor, and turbine blades can be characterised. This is exactly what we excel at, understanding the mechanics of a machine, and then deploying a measurement system to gain more information about its behaviour.
There’s a high likely-hood that you’ve worked with them too. MEMS accelerometers have miniaturised the technology to the point that for decades they have been able to fit onto circuit boards. Nearly all smartphones have accelerometers built in, as do laptops to detect free-fall and protect moving parts from shock. The reason that phones can work out which way up they are, so as to rotate from portrait to landscape, is due to an accelerometer buried deep inside, that allows the phone to use acceleration due to gravity as a proxy for orientation.
Do you have a complex instrumentation task at hand? Contact Flintmore today.
Further Reading
We’ve barely scratched the surface in this article of the use cases and practicalities of accelerometers. If your interest has been piqued, check out these further resources:
Omega mainly focuses on peizo-electric types, but does provide further detail in their signal conditioning, if you’re wondering how accelerometres fit into the current/voltage topic we discussed a couple weeks ago, check out their article:
https://www.omega.co.uk/prodinfo/accelerometers.html – it also dicsusses how to mount accelerometers.
Wikipedia has some great resources as usual for non-contentious techy stuff, including unique article for the piezo-electric type:
https://en.wikipedia.org/wiki/Accelerometer
https://en.wikipedia.org/wiki/Piezoelectric_accelerometer
In particular I would recommend the main article just to see how many use cases the accelerometer has and how widespread they are.
Texas Instruments have some fantastic resources for most instrument types, we’ve linked to them in previous articles. Whilst I’ve been able to find some datasheets and use case reports from them, I havent been able to track down their over-all guide, which is widely recommended. If you track down a free (and legal) copy, send it our way so we can link it here.
However, this datasheet on the DRV-ACC16-EVM gives an indepth example of another use-case for accelerometres:
