663 views
<center> # Vibration Mode Analysis *Originally published 2022-08-05 by Nick Sweeting on docs.sweeting.me/s/blog.* </center> --- Detect when your machine is failing by feeling its vibrations and hearing its sounds. We're luckily hardwired to be able to pick out complex relationships in harmonics, oscillations, etc. with extremely advanced hearing and touch senses. *Listen to your bicycle, motorbike, car, train, plane, etc.:* When a new sound or rattle or vibration appears, make note of it. This is some new information that you have gleaned from your engine about its internal state, and all changes in sound happen for a reason. Once you are acutely aware of the healthy frequencies and harmonics present during normal operation, try deliberately altering some aspect of your bike and listening for the change in sound. A worn out stretched chain will sound significantly worse than a new one, even to the untrained ear. However you can also detect much slighter changes like the emergence of even a single new stiff link in your chain, or a loose bolt somewhere on the frame. You can also hear the amount of tire deformation based on the tire noise, wheel slip based on gravel noise, etc. All the wide range of frequencies expressed in your bike are usually harmonics of a much smaller set of individual root notes. Engine rpm is the obvious one. The throbbing you feel in your foot peg is likely in sync with the piston firing, or perhaps with a flywheel defect causing imbalance, but either way its always tied to engine speed. Separating the frequencies can help you isolate a fault. For example if trying to discern whether a new rattle is coming from the chain or the transmission, disengage the clutch and see if the sound decreases in sync with the engine rpm or tire rotation rpm as they start to diverge. Every independently oscillating part (pistons, flywheel, tires, etc) has the potential for creating a new separate fundamental and a whole swath of harmonics as it resonates other parts in your bike. When a part begins to fail, it will emit sound in sync with one of those fundamental frequencies, depending on which subsystem of the bike it's a part of (wheels, engine, chain, steering, etc). Each subsystem has a root note and harmonics that modulate together. From relative frequency relationships alone you can usually isolate the location of a new fault before you find it visually. If different parts of your bike are oscillating or spinning at unrelated rpms, all their upper harmonics from other vibrating parts also become clearly unrelated. allowing for separate fundamental notes to illuminate which harmonics are separate as well. This allows you to do a general auditory health check of the entire bike while riding, and if you hear a fault you can efficiently binary search to find it. The trick is to--one at a time--separate the spinning rpm of each subsystem (e.g disengage the clutch and slow the wheels with the brakes) and listen for unhealthy frequencies that shift with that new modulation. This usually is a strong indicator they must be a part of that subsystem. A clicking noise coming from the chain may abate or strengthen depending on how much you load the throttle, but it can also reveal how stretched the chain is as average slack significantly affects chain noise pitch and tambre. You can also usually isolate noises that involve liquid from those that don't, e.g a failure in engine case, coolant system, brake reservoirs, gears submerged in liquid will usually sound "liquidy". Bearing failure sound is unique from metal fatigue cracking sound is unique from cavitation sound, etc. I want to release an FFT display console that reads from microphones that attach all over the bike. It highlights each root note and group of harmonics in a unique color, and highlights when new ones appear with an alert notification and prediction of origin with a basic pretrained model. The user could also train the model with specific known failure noises for their bike. Expensive industrial systems exist to do vibrational mode analysis for aviation. Automotive, etc manufacturing and design, but I have yet to see an affordable ip7x rated one for personal vehicle road/dirt/marine use. A more general statement can be made here too: learn the patterns of the machines in your daily life, and observe the nuances in regularities and irregularities as you use it. Learn to group and isolate symptoms by their smaller set of possible underlying causes. Push the limits slightly when learning and calibrating a new system to calibrate what the operational limits sound and feel like. If we step back even further we could apply this to human relationships. Pay attention to the small daily habits and feedback routines we give each other, because a faltering in those daily heartbeats could indicate deeper problems. ### Links https://store.ncd.io/product/2-channel-industrial-iot-long-range-wireless-ultrasound-vibration-sensors/