I want to make a network of cheap(er) strain gauges or load cells to improve accuracy to measure a torque between 1 and 3.5 Nm by means of a string or chain attached to the shaft of an emotor.

1) Is this a valid life goal?

2) Currently assuming a simple parallel network is the best option. Is this accurate?

3) Assume force measured N, number of gauges is n and "midrange" of each gauge is (gauge_max - gauge_min)/2 . Further assume number of gauges is determined by budget. Is the best option for each gauge;

a) N = midrange

b) N = midrange * n

c) something else?

Many thanks in advance for any help

Joe

 

if you did that, wouldn't you spread the strain across those paralleled sensors? Like 3 sensors would all read load/3. I think in this case you would be introducing more error instead of less.

 

Generally, averaging the results over n sensors each with a normal distribution of accuracy will improve the overall accuracy by √n.

 

Consider that in many processes the production deviation from "perfect" is not a random distribution, but rather caused by some "offset deviation", it is not likely that the ultimate accuracy will be improved. Production runs of stable processes tend to not produce random deviations.
The TS has an interesting theory, but they need to have a much better understanding of the process.