Robotics Case Studies
Large Cement Manufacturer Sought Faster, More Accurate Solution to Quality Control
by CrowdANALYTIX Content Team
CrowdANALYTIX Inc Posted 09/02/2020
Inconsistent quality proves problematic
In the United States, the construction industry employs more than 7 million people and generates well over $1 trillion every year. The manufacture of cement is one of the key industries that supports construction, since cement is needed for a high percentage of new structures as well as for repairs, maintenance, and elements of the American transportation system like roads.
Cement manufacture is a complicated process in which limestone, clay, and/or shale are extracted from quarries, crushed to powder, and then blended in specific proportions intended to ensure quality and integrity. However, the current means of quality testing requires lab analysis and therefore can’t be performed very often during the production of mass amounts of fresh cement.
Testing methods could be sped up
A large cement manufacturer decided to seek a new means of quality control, namely a more precise prediction of a single key quality parameter: the amount of Free Lime present in a batch of cement. If the amount of Free Lime could be accurately measured, this company thought that they could achieve:
- Reduced heat consumption by adjusting the process in the cement kiln through prediction of Free Lime (better energy efficiency).
- Improved production rates.
- Improved quality stability and compliance.
Achieving these goals could lead to higher quality cement, faster production, and therefore greater profits as the company improved the supply of cement to their broad base of customers.
CrowdANALYTIX was ready to take on the challenge and attempt to improve the process of cement quality analysis. We utilized more than 20 teams comprising over 400 professional solvers to address the problem. They used multiple prediction engines to analyze a wide range of product samples containing Free Lime, measuring the amounts present. Data was available across four years, from 2016 through 2019.
Two of the most optimal models were picked for final deployment. These models 1) met both precision requirements and 2) provided the most explainable predictors that could be used to improve the quality process. The algorithms were then deployed and made available across cement manufacturing plants, thereby improving the quality of the product and optimizing the manufacturing process.
In early field tests, production rates were expected to increase by over 15%, while reducing energy costs by 9%. This is predicted to lead to millions of dollars in savings for the cement manufacturer.