Sensor Selection for Measuring Timing

[TugboatEng]

I have an idea that involves using sound to detect the occurrence of certain events in a diesel engine. Specifically, I am looking to "hear" the rocker arm contact the valve stem, the valve closing in the seat, and the opening or the fuel nozzle. I would like to be able to do this through a thin die-cast aluminum valve cover. The ability to measure the frequency of the sound may be beneficial in identifying which event is occuring. However, I will know about when each event does.

The question: Are there any recommendations for sensors to start with?

 

[IRstuff]

You just need basically a contact microphone with a high enough frequency digitizer and the rest is signal processing.

 

[GregLocock]

You might find an accelerometer on the head is more useful. It's a tricky project, all those events are at the same fundamental frequency, they'll just have different phases and harmonics. A stethoscope or a screwdriver might be a good place to start.

 

[TugboatEng]

I will know approximately when each event is going to happen within a few ms. I have an individual head per cylinder so that should also help with isolation. I would like to be able to make this measurement for each cylinder individually with a semi-portable tool.

These measurements will be correlated with a crankshaft position sensor.

 

[GregLocock]

If you are confident with the timing you can use a gated time window to get rid of the noise. That is you zero out everything but the few degrees of crank sensor around that event.

et voila

https://www.bkvibro.com/wp-content/uploads/2020/12/BAN0062EN11_MS_Gated_vibrations_recips.pdf

 

[TugboatEng]

The goal is to verify that the timing is within a window.

That article is very close to what I want to do. I am interested in the best type of sensor to start with. I operate many of the same type of engine and think that I can do some empirical analysis to generate a baseline and use any deviation to trigger additional inspection.

 

[IRstuff]

For truly synchronous events, i.e., those with know timing, boxcar averaging can work wonders; it used to be claimed that you could acquire signals that were SNR~1 and beat the noise into submission.

https://en.wikipedia.org/wiki/Boxcar_averager

 

[TugboatEng]

Specifically I am trying to measure the duration of valve opening by listening for the rocker contacting the stem and the valve contacting the seat. I'm not interested in the specific frequencies of the sound, yet. I am just looking for spikes in amplitude and I know approximately when these spikes are going to occur which should make them easier to isolate from the rest of the engine noise.

 

[IRstuff]

The box car averager still sounds like the way to go, but, all the engine noise is likewise synchronous with the valve opening/closing sounds.

Websearch seems to suggest that pressure changes in the cylinder is the more typical way of detecting valve operations.

 

[TugboatEng]

On larger slow speed engines there is a pressure based measurement called a weak spring diagram which does show valve timing. I am working with high speed engines that don't have test cocks for cylinder pressure monitoring.

Boxcar averaging aside, I don't have a sensor to make a measurement yet. I have some bolt holes in the heads I can attach a sensor to. I would really like to be able to hold a sensor on the valve cover directly above the valves like a stethoscope. I don't know what type of sensor to start with. I need data before I can start filtering. Should I be looking at microphones? Accelerometers? I know I'm going to have to go through some iterations to get to an ideal solution but I don't even know what to Google yet. I acknowledge that I may need one sensor to detect opening and one to detect closing.

 

[IRstuff]

A decent microphone has bandwidth beyond 15 kHz, while the accelerometers for vibration testing can hit upwards of 20 kHz bandwidth. There are piezoelectric accelerometers that supposedly reach 90 kHz. which might be consistent with assuming 6000 rpm --> 100 Hz, possibly in the range of 5 deg crankshaft timing resolution, so 7 kHz event resolution, so 70 kHz edge resolution.

 

[GregLocock]

Knock sensors work 5-10 kHz, they are detecting high frequency modes in the cylinder head. Finding where to put them is a bit of a project on an automotive engine because we are cheap and only want to use one sensor to detect knock in all the cylinders. Since you've got single cylinder heads you don't have that problem as much. If you want to use a mic you can get ultrasonic mics, 40 kHz would be pretty easy. https://zachpoff.com/resources/cheap-microphones-for-ultrasound/ I'd start with one of those.

70 kHz is a bit more spendy https://sonorousobjects.nyc/product...1n2mrH9KzkEFptCUZS-032JaOwFw6oDNzjtY5ME4vsmTj - still pretty cheap in the world of measuring instruments.

For added excitement you can also use a pair of mics to create a directional probe, this is called acoustic intensity. This will improve your S/N ratio by filtering out the noise from the other cylinders.