California Scientific

California Scientific
1000 SW Powell Ct
Oak Grove, MO 64075
California Scientific  *  BrainMaker Neural Network Software  *  Predict Forecast Classify Stocks Bonds Markets Commodities Diagnose Medical

BrainMaker Neural Network Software

Forecasting Required Highway Maintenance with BrainMaker Neural Networks

We've all driven on a road that is full of pot holes or cracks. You can barely hold your commuter cup and youre anxious to get around that big semi so you can get into the smooth lane. But then you ask yourself, didn't they just fix this road last summer? Chances are you're right. But experienced highway maintenance engineers are hard to find, and as a result, the appropriate treatment isnt always selected.

Professor Awad Hanna at the University of Wisconsin in Madison has taken the guess work out of the maintenance and repair process by training BrainMaker to become a maintenance expert. If a seasoned professional isn't available, a recent college graduate and a computer program can do the job with a high degree of confidence. Since there is no mathematical formula to solve this kind of problem, its an ideal application for BrainMaker.

Professor Hanna trained the neural network with information provided by experts who can tell with a high degree of accuracy (confidence) which type of concrete is better than another for a particular problem. These experts were given a variety of situations and asked to provide various treatments. Professor Hanna then trained using the back propagation method on 1 hidden layer. Currently Professor Hanna is developing a simple program to be used with BrainMaker that will take the input from the user and produce the most appropriate output based on previous experience provided by these senior people.

Some of the inputs include qualitative values for temperature and volume of a particular piece of pavement. Due to lack of funds, the number of input values was limited to 10. The output is the pavement treatment associated with a degree of confidence. For example, the recommended treatment might be chip seal with a confidence of 8 out of 10. Because there are so many variables, rarely is there a situation that occurs with 100% confidence.

While Professor Hannas research is focused on a Midwestern area that experiences cold, ice and snow, and is based on the input of experts from this area only, his methodology could be applied to any geographic location. If human experts are not available to provide input, routine maintenance data from any Department of Transportation can be used instead. According to the Professor, Usually there is some kind of historical record of work that has been done in particular sections of road over the last few years as well as an annual evaluation of the Riding Comfort Index.

The Riding Comfort Index is a rating of how comfortable you are on a particular section of the highway. The smoothest, best road would score 10 out of 10. A bumpy road for example would score a 5. Roads are measured before treatment and again a year later. If a year after treatment the road is still scores high it means the treatment was a good one. If it the score is low it means the treatment wasnt really appropriate.

To validate his results, Professor Hanna is seeking funding to test his program. His testing will include more highway locations, different types of cracks and bumps, and data collected at different times of the year. He plans to record various sections of a particular highway (For example I 94 between miles 101 and 102); classify the cracks found there (severe, medium or light); look at external factors (annual daily traffic, air temperature, amount of snow removal, wind factor); input all facts into the software, and obtain recommendation for the ideal treatment.

BrainMakers transfer function will be used to determine the exact confidence -- based on information provided in the training phase.


10 different inputs, including:
qualitative values for temperature
volume of a particular piece of pavement


Appropriate maintenance treatment and degree of confidence.