Pushrod strain gauges

As I have a thing for measuring things, this summer I did an attempt to put a strain gauge on one of the push rods. A strain gauge is like a thin film with an embedded wire. You can glue it on objects and when a force is applied to the object, the electric resistance of the wire changes and from that you can measure the strain. Strain gauges has many applications, such as measuring the weight on a crane or the torque on a shaft. Or for the electronic bathroom scale.

At first the plan was to glue the strain gauges on the springs instead of the push rods. I bought one car spring with the lowest price and lowest spring constant and started experimenting on the kitchen table. I used a cyanoacrylatec glue (like super glue) for the gauge and put quick epoxy on top as protection.

Strain gauge pair glued on a spring

OP amplifier experiment

The resistance change from a strain gauge is very tiny so it has to be amplified a lot. I used a Wheatstone bridge and a 100x amplifier. It worked as a charm. The output reacted nicely even with small loads on the spring.

I drew a small circuit board and ordered it from a Chinese PCB manufacturer. Amazingly easy these days, and costs like $15 for five boards. One board contains two amplifiers so two boards would be necessary for logging data from all corners. I used to work with doing things like this, but it was a long time ago and my soldering skills are not what they used to be. Neither are my eyes.

PCB board with dual amps, bridges and power regulator.

I decided to do a test by putting the strain gauge on one of the push rods instead of the springs (see the top picture) and wired the output to my data logger.

Steering back and forth on the road.
The lower signal is from the strain gauge.

It worked!

Still, as the push rods are quite thick the signal was very weak and contained lots of noise compared to the signal I got from the springs.

Now the question was, what to do with all this new data? 

One of the goals was a way to measure aerodynamic lift. The other was if I could find any undamped suspension frequencies. I took a drive on the motorway when there wasn't any traffic.

I started to play with Matlab. I don't know much about signal processing, or Matlab, or even maths in general. Only having one signal instead of four also made things harder. I did what I could to compensate for weight transfer using the logger's accelerometers and got this graph which with some biased imagination could confirm the front end lift. 

Speed vs front downforce (in this case lift)

I also tried to do some FFT analysis but couldn't get any interesting results from that.

Then I took the car to a track day.

Blue = track day, red = road

This is a histogram of the strain gauge signal derived over time (ie the rate of the signal change). The red bars are from a calm road test, the blue bars is from a track day session.

High end data acquisition systems (such as Motec) have this kind of analysis functions built in for suspension travel sensors. The theory is the histogram should ideally follow a certain shape, and be symmetric along the centre. As this signal is from the pushrod and not suspension movement, I assume a low rate of change is a good thing as it reflects the tyre's contact patch. I don't fully understand what conclusions I can draw from this.

So what have I learned?

Well, not very much new from the gathered data. Also that I'm good at starting projects and quite bad at finishing them. But I knew that as well.

Still a very fun project.