USE CASE: CONCRETE

A concrete recipe contains the list of ingredients that provide a particular concrete product its resistance and other performance metrics. The performance metrics are usually measured 28 days later than production. These ingredients are cement, water, additions, additives and arid aggregates, where cement is the most expensive ingredient in a concrete recipe, representing near to 90% of the total ingredient costs.

THE AI SHOULD PERFORM TWO TASKS

  1. Correct cement overdosage: By assessing if a particular recipe of concrete (list of ingredients and their quantities) has the potential to reduce its cement usage and still achieve its required resistance defined for a particular product, and if yes, then propose a change in the proportion of its ingredients in order to reduce the amount of cement used and still achieve the required resistance.

  2. Correct cement underdosage: By assessing if a current recipe will not be able to achieve its required resistance and if detected, then propose a change in the proportion of ingredients in order to achieve its required resistance.

droide comandante
Standing Meeting

TESTING, DEPLOYMENT AND IMPROVEMENT OF AI SOLUTION

Testing of AI results will be done by comparing the proposed corrections in a test dataset of known historical results (but unknown to the AI) and compare the proposed changes to the actual results obtained. This test is not complete as it will correctly test whether the AI can detect under or overdosage but will provide only an approximate result about whether the amount of the correction is enough to achieve the required performance, since the historical data does not contain the corrected proposed amount of ingredients.

USING THE ALGORITHM

A given finished product code (concrete) has a recipe just adjusted by engineers following formulas contained in their “Handbook of Dosage” considering the latest quality control results of the ingredients to be used. That finished product is then ready to be produced and loaded in the Information System of the Plant.

But before proceeding with the mixing of ingredients, the AI assess each recipe and come up with corrections to the quantities adjusted by the engineers. The engineers review the changes proposed and can either accept, modify, or reject the changes proposed. A log is generated in order to further improve the recommendations.

Electricista con tableta
Cement

BILL OF MATERIALS

The AI analyzes the list of recipes ready to be mixed at the plants and the
expected resistance of each product to be produced and proposes changes in the
proportion of ingredients in order to correct under and overdosage of cement.