Reinforced Concrete Service Life Prediction with Neural Network Analysis

I applied a neural network program to concrete service life data and used it to predict the useful service life, and it worked, almost, maybe.

Applying the Model

I found a model that I could interpret online and then modified it where needed to work with my data sets. Finding data was a mission as data is valuable and not everyone puts it all out there. Then you also need to limit the variables to test viability, so you limit the studies you choose and basically just limitations on limitations.

Training the Model

It took a long time to run sometimes, depending on how much I wanted it to do. There was a correlation with more training giving better results, but I didn’t put in the time to find the point of diminishing returns on accuracy.

The Results

It worked, kinda. It gave predictions very close to calculated values, but the variables were limited, and thus I think it made it easier for the model. But this means…..

Where Next?

I hope someone does something more on it. I don’t want to take it further as I don’t see much commercial viability for the project, so I’ll leave it where I leave it.