RoverGauge Data Analysis

[blaze_away]

Having now helped a number of people over the last 3 years with cars running rather badly on the original 14CUX system I am now ready with enough data to confidently publish and say that bad MAF sensors can be identified by logging RG data at idle and then doing some statistical analysis of the MAF output.

The data needs normalising first (method that statisticians use to be able to compare data so that they are comparing apples with apples.) In our case engines with different idle target rpm, maps, injectors, plugs, timing etc, ie all the things that make an individual engine unique in itself.

Once that is done then analysing the variability of the output of the MAF sensor will reveal the MAF's health.

On the chart below (the lower chart) I have ranked the MAF's in term of there health, The upper chart shows the average (mean) and range of output of each maf, this varies according to the type of maf, engine size etc so will vary from car to car. The mean seems to be not important as the 14CUX learns to compensate for this via the long term trim adjustment.

5 on left are all good, showing good mpg, engine response, lambda control and plug condition.

2 in the middle were a bit suspect showing marginal performance poor running, blackish plugs and lambda not always in control.

3 on the right are really quite bad with engines showing really poor mpg, black plugs, usually lambda's maxed out and poor running overall..

In the case of the 5 marginal/bad maf's once they replaced them with either good used items or a new one (from reputable sources) they resulted in producing the 5 good maf responses.

In summary a good MAF will have a very tight output at idle with standard deviation of 0.6 or less (the lower the better) whereas a bad MAF will have standard deviation of 0.7 or more, the higher the number the worse the engine runs.
Look forward to comments and critique on this from you knowledgeable folks.

[geordiepingu]

Hi Frank

I'd be keen on knowing the details of how your normalisation and analysis is performed. It is likely that I could build a tool in my downtime to use a 14CUX diag cable to automate the testing of this and do it right on a computer now there's a reasonably proven method for doing this. Would be much quicker than datalogging and then doing post analysis.

Thanks

[blaze_away]

geordiepingu said:

Hi Frank

I'd be keen on knowing the details of how your normalisation and analysis is performed. It is likely that I could build a tool in my downtime to use a 14CUX diag cable to automate the testing of this and do it right on a computer now there's a reasonably proven method for doing this. Would be much quicker than datalogging and then doing post analysis.

Thanks

Process is quite detailed, I have it in pdf but can't post it here as it just takes image files, drop me a msg with your email and I will fire it over to you.