Al enhanced concrete batching
Someone once told me that if I forecasted the weather for tomorrow to be the same as today, I'd be at least as accurate as the meteorologist on TV. True or not, we know accurately predicting the future is difficult. People store memories of challenges and experiences then based on recollection of those experiences we make predictions. Some have better judgment than others but without a doubt there is no substitute for experience.
A limitation is human memories are imperfect ... we forget details and rely on feelings or related, but not identical, experiences.
What is artificial intelligence anyway?
Artificial intelligence (Al) is the latest buzz word ... like fuzzy logic, 3D printing, or quantum computing it will save billions of dollars or countless hours in our daily lives. Some maintain Al is the beginning of a new age ... smart machines to replace people in dangerous or undesirable occupations.
But that is not here yet. How will applying Al-enhanced concrete batching help you today? To understand that let's first talk about how Al works. There are two out of the possible four types of Al that are in use today1 . Types three and four are self directed, thinking machines that remain science fiction.
Today's Al is based on 'artificial neural networks' or ANNs. These ANNs are just (really, really) powerful and fast educated guessing machines.
Using artificial intelligence do something useful.
Let's return to weather forecasting. It's the year 1850 and the available tools include a thermometer, a barometer, wind speed/direction, and visual observations. With such limitations your forecasts are hit or miss, you can't know over the horizon a cold front is coming. Fast forward to the 1920's where you have telephones and radio. The meteorologist can access more information than before and makes better predictions than his predecessor. In the year 2024 you get thousands of data sources from the Internet.
A human meteorologist with years of experience can look at some number of these reference points and divine a good guess of what the weather is going to be. But this is still a person, with imperfect memory using a limited number of data points in the 30 minutes before they must produce the next forecast. This is where the ANNs come into the picture, just as applying Al-enhanced concrete batching will benefit our industry.
How artificial intelligence can improve weather forecasts (and concrete batching!)
Our meteorologist provides updates to the forecast using experience and data inputs described earlier. Humans use experience, other people's opinions, gut feel, and anything else they can come up with to make the best estimate possible. Recall that our memories are frail. .. imagine you had a photographic memory that remembered every detail and outcome ever experienced in every situation encountered ... and that information was instantly available to you ... that's the potential of an ANN.
An ANN 'reads' some number of data points, or inputs, and evaluates that data against what has happened in the past. Over time it acquires 'experience' (comparing the inputs to the results) and the ANN 'learns' what is most likely to happen when presented with a certain set of data; certain patterns of inputs statistically create certain patterns of outputs. What makes an ANN possible is it has practically limitless memory, infinitely more capacity, and can perform millions of comparisons in seconds as compared to humans.
For our weather forecaster, if their ANNs were provided thousands of inputs from global weather stations, with the actual weather from thousand of data points around the globe, and this was done over time the ANN will be able to correlate the current weather measurements (wind direction/speed, atmospheric pressure, temperature) from each point and make a pretty good prediction as to what was going to happen next. It won't yet achieve perfection because there are random events that are not available to the ANN (i.e. sunspots, volcanic eruptions, or a butterfly flaps its wings) but it will get very close. The rule is, the more inputs to the ANN, the better the prediction (forecast.)
Examples of Al-enhanced concrete batching
Al is the ability to look at data inputs, process that pattern of data then makes a prediction as to what the result will be.
The ready-mix plant automation system is feeding materials from the overhead bin/silo into the scale. The ANN (such as Command Alkon's Mlntelligent Feed) is constantly reading the current weight on the scale, the flow rate, and perhaps air pressure or ambient temperature. During its training the ANN has monitored hundreds of similar feed cycles. It has 'learned' that given a certain scale weight, flow rate, the material being feed, and other inputs, it can predict the results of the next few seconds with high confidence and therefore 'predict' when to close the gate without going over tolerance. (The more experience the ANN has the higher the confidence will be.) This mean that the automatic batching weight up process can be more accurate than ever before.
This is not magic, nor is it true 'intelligence.' It is a powerful brute force tool, made a reality through the vast computational power in the cloud and high-speed interconnections to/from it that will vastly improve the production process. The ability of an ANN to continuously learn and adapt means the manual 'tweaking' of plant parameters is obsolete!
This is one example of how Al-enhanced concrete batching can improve the production process. If a process is such that sensors measure the inputs to the process, and sensors measure the results there are boundless opportunities to use Al for improving the margins of our business.
Up next: Tips for reducing balling in concrete.
As a key architect behind COMMANDbatch and REDI-VIEW, Randy Willaman brings decades of hands-on experience with concrete batching systems. Through his posts on the MARCOTTE Blog, Randy shares his unparalleled expertise and insights with concrete producers looking to maximize their system's potential. Learn more about his services
