Engine Vibration Query

[system 4-50]

If I mounted an accelerometer and microprocessor in a rigid package and secured it soundly to my engine block and recorded the movements of the engine with respect to the hull, could this give useful information?

I'm wondering if it could tell me if I have diagonal rocking, bow-stern rocking, or cork-screwing?

Would this help to check that the drive alignment was correct?

Any ideas on the vibration frequencies of an engine, in this case a Beta 43? If they are too high I would not be able to detect them accurately.

(I like using microprocessors to record performance data, it irritates my girlfriend somewhat...)

[Tony Brooks]

I suspect that all depends upon how sensitive the accelerometer is and how its design. Certainly I doubt one that uses a weight against a number of Pietzo elements would work because of the vibration frequency but then my automotive type knowledge is way out of date..

 

I suspect it may require strain gauges between the engine and hull so you are measuring relative movement and not any movement imparted on the engine by the hull.

 

Even if you do measure any movement I am not sure how you are going to minimise it or use it.

 

Hold onto that girlfriend

[OldGoat]

A solution looking for a problem maybe??

If you had an extra heavy pre WW1 lump of an engine then some analysis might be "interesting".

BUT with a modern lightweight engine I doubt whether any measurements could tell you anymore than could be obtained by a visual inspection.

 

Is the engine fitted with vibration mounts - if you waggle the engine from the top does / do any of them feel floppy

Do you have a "proper" engine coupling - quite often they are omitted or 'not the best type'.

[Paul C]

I suspect you'd need to obtain a "baseline" of normal vibration, ie well aligned, engine mounts in good condition, firing evenly on all cylinders. Then you could do further test runs with a fault such as 1 cylinder not operating, misaligned, engine mount failed etc to get data on the pattern with the fault. Trying to work out the particular fault from raw data would be a nightmare.

[system 4-50]

Beta 43

Vibration through the hull much greater than 3 years ago. Then I never noticed it.

1 mounting replaced a couple of years ago after getting soaked in diesel.

Centaflex

Standard mountings (like K47??. Not sure I remember the number right.)

Have obtained some replacement mounts & will be changing them when I can get it high enough up the priority list.

Considering the replacement job, I just got to wondering thats all.

 

Thanks for the comments.

[nicknorman]

Any ideas on the vibration frequencies of an engine, in this case a Beta 43? If they are too high I would not be able to detect them accurately.

 

It's an inline 4 so secondary balance is poor, ie most of the vibration will be a twice rotational speed. So at say 1500rpm that would be 50Hz. See here for info on generic engine vibration: https://en.m.wikipedia.org/wiki/Engine_balance

 

Of courses added to the reciprocating/rotational vibration of the actual engine itself you would have the natural frequency of the engine moving in the flexi mounts, the different frequency associated with the prop shaft (presuming the gearbox was not 1:1), possible additional stuff caused by any UJ type couplings, and all that interacting in a way which varied according to engine rpm.

 

Out of interest we had vibration monitoring on helicopters with complex drive trains that took an engine rpm of 23,000rpm and reduced it to 265rpm by means of various gears from spur gears at the fast end to bevel gears in the middle to epicyclic gears at the slow end.

 

It worked thus: there was a timing reference on the high speed end (ie 1 pulse per rev). To analyse a particular gear or shaft, signal averaging was used, ie data was recorded for 1 rev of that component (knowing the gear ratio, x revs of the high speed end) and then the next rev was overlayed onto the first's data and so on for perhaps 100 revs. The effect of this signal averaging was to reinforce the data of that shaft (ie it was correlated) and since the data from other shafts at other frequencies was not correlated, it tended to wash out to low level mush. Ie the signal averaging allowed the signal from one shaft inside a gearbox to be picked out from all the other shafts, using an externally mounted sensor that could "see" several shafts.

 

Then the signal averaged data was run through an FFT to put it into the frequency domain, of course you would get the primary (1r) frequency and some gear meshing frequency depending on number of teeth etc, all of which was expected. So in the frequency domain you would subtract the known and expected frequencies. You then had a much smaller signal that you ran through an inverse FFT to put it back into the time domain and what was left represented the "unexpected" stuff so that for example a cracked tooth in a gear could be identified as a pulse at some specific shaft angle.

 

Well all that is a bit OTT for a boat but it just goes to show what you can do with some fancy digital signal processing on what looks initially like a load of noise.

 

Edit: don't forget Mr Nyquist who said you have to sample data at at least twice the frequency you want to measure.

Edited April 15, 2016 by nicknorman

[David Mack]

Beta 43

Vibration through the hull much greater than 3 years ago. Then I never noticed it.

1 mounting replaced a couple of years ago after getting soaked in diesel.

Centaflex

Standard mountings (like K47??. Not sure I remember the number right.)

Have obtained some replacement mounts & will be changing them when I can get it high enough up the priority list.

Considering the replacement job, I just got to wondering thats all.

 

Thanks for the comments.

 

 

Even if that one mounting was the identical model to the one it replaced, it is likely that it stiffness characteristics will not be the same as the other three which will have aged. And in replacing it did you get the engine properly aligned and balanced anyway? I would have thought that a likely cause of some unbalanced engine movement.

[system 4-50]

 

 

Even if that one mounting was the identical model to the one it replaced, it is likely that it stiffness characteristics will not be the same as the other three which will have aged. And in replacing it did you get the engine properly aligned and balanced anyway? I would have thought that a likely cause of some unbalanced engine movement.

It is possible that this is so. I didn't do the job myself, I got a professional in to do it.

[Chalky]

There are some really good micromachined accelerometers out there that aren't that expensive and can give really good results. A lot of them are designed for robotic navigation (can give really good results with an extended Kalman filter) however they may not have the bandwidth you're looking for. I bought a 6DOF IMU with I2C interface off ebay for about £3 and a 9 DOF one isn't much more. You could get far more interesting data than just engine vibrations!

[Tiggs]

nicknorman said most of it. A usual technique is to you 3 accelerometers on a block giving X, Y, Z directions, Start the engine and run it very slowly up to max speed (taking about 4 minutes to pick out any resonant frequencies) then take a similar time to decal down to idle. FFT this and plot it against speed (you need a speed signal for this recorded with the vibration data). The whole thing gets really too complicated and you are better off aligning the engine properly on some decent mounts.

[Iain_S]

(snip)

(I like using microprocessors to record performance data, it irritates my girlfriend somewhat...)

 

Well. if you can record your girlfriend's performance with microprocessors, an engine should be easy

[dmr]

Strictly you will be measuring the absolute vibration of the engine rather than the vibration relative to the hull, but as long as engine vibration is much bigger than hull vibration that's neither here nor there.

Choosing a good location and orientation for the accel. will be important (to avoid the big vertical component that Nick refers to) but if you do it right and get good baseline results then you can derive some useful information. This is classic machine health monitoring. You should be able to spot mount degradation and even developing imbalances between the cylinders.

Correct accel. mounting and sampling is important.

A TDC sensor of some kind will make things a lot easier and more interesting.

 

If you go ahead with this then keep in touch, I used to do this sort of stuff a lot and still do a little bit.

Main frequency will be 2E (twice rotational) plus a fair bit at 4E and 6E etc. Anything at 1E and 1/2E will indicate potential issues or interesting things going on.

 

................Dave

Edited April 16, 2016 by dmr

[system 4-50]

Thanks dmr. Will do. Unfortunately it will have to wait until my gas installation is complete and one or two other things get done.

[dmr]

Thanks dmr. Will do. Unfortunately it will have to wait until my gas installation is complete and one or two other things get done.

 

Don't worry, I have a lot of projects that never quite make it to the top of the actually getting done list, but its still interesting to think about these things.

Depending upon what frequencies you want to get up to the accelerometer mounting can be a real issue. Ideally a proper accel well fixed to the engine would be best, with a separate box of electronics. These are usually not cheap. The circuit board mounted jobbies should get the really low frequencies but the bending and twisting of the circuit board itself will soon be an issue. Its surprising how much these things move. One little job I did was looking at the movement of circuit board and electronic components inside one of these ECU's that the makers of big diesels like to bolt directly to their engines.

 

................Dave

[system 4-50]

Mustn't.

Mustn't.

Mustn't start yet another project before I've cleared at least one existing one off the in-progress board. No matter how intriguing..