Artificial Intelligence boost for the Cement Plant

AI and a cement plant

Published by FirstAlign

Today, Artificial Intelligence (AI) is commonplace. Siri, Alexa, Netflix, and SatNavs are, to name but a few, of the AI applications that millions of people use daily. AI has made significant changes to supply chains and administrative functions, but what about large scale production industries?

Well, AI’s presence in production has been scant, but not without momentum. One such industry that is adopting AI is the cement manufacturing sector. The challenges faced by the cement industry are many. To name a few; high energy consumption, rising costs, process complexity that is inherent to the industry.

There are other global concerns such as environmental deterioration, lack of natural resources, greater rivalry in the global market. To take on all these challenges the cement industry requires greater operational efficiency and better fault prediction systems.

For a long time, cement companies have been digitizing plants with distributed and supervisory control systems. However much progress has been made in analytics and decision-support. Operators largely rely on their experience and intuition to make decisions.

For Example, control room operators manually monitor numerous signals on-screen and adjust settings accordingly. They must also troubleshoot and run tests at the same time. As a result, many operators must prioritize activities on urgency, and not necessarily the important/ high-value ones.

Frequent adjustments that rely on human judgment results in higher than necessary energy usage and greater wear and tear of machinery. Also, reliance on experience for decision making can prove problematic when it comes to resource management.

Artificial Intelligence’s capability to standardize and improve knowledge can eliminate the need for dependence on people alone. AI can make complex operational decisions on its own and can provide better support than conventional decision-making technologies.

The cement industry is highly energy-intensive. In the US, 349.4 trillion British thermal units of energy were consumed by the cement and lime industry in the year 2019. It is expected to reach 360.1 trillion British thermal units by 2050. Cement plants are looking for innovative ways to reduce energy consumption and costs.

Superior performance, high processing power, and cheap memory paves way for versatile AI solutions that are more adaptive to industry needs, assisting in fully automating complex tasks. It requires less manual power to maintain and can be adjusted to revised strategy and production plans.

Why Artificial Intelligence for cement plants?

There are multiple potential applications, such as;

  • Failure prediction (operative and corrective failures);
  • Production processes optimization;
  • Predictive maintenance;
  • Remote operation; and
  • Product design and quality; smart supply chain.

Working of a cement plant

Cement manufacturing involves three major steps;

  • Firstly, limestone from the quarry is crushed. Raw materials such as iron oxide, aluminum, etc. are mixed with the crushed limestone and ground to produce a raw meal.
  • Secondly, this raw meal is passed through a cement kiln at high temperatures to produce a clinker.
  • Finally, the clinker is ground at appropriate temperatures to produce cement.

How a global cement company adopted AI solutions?

In March 2019, Cemex, a global building material company agreed with Petuum, a US-based Artificial Intelligence company to implement Petuum’s Industrial AI Autopilot software products.

Industrial AI Autopilot

The Industrial AI autopilot refers to Machine Learning and Deep Learning that helps to control complex processes to obtain better results. It helps operate at a higher level than a human operator. Human operators rely on their experience to make operational decisions, which need not be an efficient option.  An Artificial Intelligence led system can operate at a much efficient range. It opens for increased process efficiency.

To do this, Petuum uses deep learning neural networks. These are a large matrix of inputs and timestamps from cement processes. In the case of Cemex, two years of plant information was fed into neural networks.

Modeling for decision making

It then modeled the relationship of variables with each other over time. Based on historical operating models, assessments of and the future course of processes, can be decided.

Modeling for optimization

Optimization algorithms create an AI-based model that provides optimal settings. These settings are recommended for the plant. Thus, Artificial intelligence analyses data of the plant to optimize and predict what processes should do. For instance, AI can predict set points for control variables in real-time. Operators validate if the recommended set points are within the range before taking the final call on the control settings.

AI Autopilot interaction with a cement plant

The AI Autopilot product is a software service that integrates with process data history and data infrastructure. It feeds on historical and streaming data, based on which it makes predictions. AI Autopilot is integrated within the control system to apply these prescriptions directly under Autosteer mode.

AI Autopilot Architecture
Autosteer work mode

The prescriptions made by AI Autopilot are posted to the plants’ real-time infrastructure, from there data is sent to the control system to be acted upon. On a five minute rotation, it triggers, ‘allow the regulatory control to move to the prescribed setting’. The operator is looking at the information and results. In a sense, now the operator is monitoring the Autosteer function.

This provides the operator flexibility to do other things. With AI, there is now more time at hand to handle more important and urgent work. Autosteer aligned with operations such as clinker cooler, preheater, pyro, ball mill, etc help lower energy consumption. It optimizes the fuel mix and maintains stable operations that result in high-quality products. 

Use case: Clinker Cooler Optimization

Phased development

Phase 1 – Forecast prediction in real-time: In this phase, output variables are identified for the AI model. It predicts behavior 5 to 15 minutes in advance. Forecasts can predict values and change in slope.

Phase 2 – Prescriptive Recommendations in real-time: AI recommends setpoints for the control variable in real-time. Operators validate setpoints, if they are within the operating range and make decisions. Kiln operators accept and manually input the prescriptive recommendation into the control system.

Phase 3 – Auto-pilot operation of kiln’s cooler section: AI mode submits setpoints for control variables to the control system in real-time. Operator monitors if autopilot operation is aligned to the normal operational range. Kiln operators can engage or disengage the control system, in case of disruption. 

When Autosteer picks up driving the asset operation, immediate results are achieved.

Benefits of Industrial AI

  1. Increased yield: AI implementation results in over 2% higher yield. Autopilot operation of the cooler, rotary kiln, preheater, etc leads to increased operation and equipment productivity.
  2. Reduced costs: The AI Autopilot model for cement can achieve 2-5% savings in energy. It cuts energy costs by minimizing energy consumption with access to real-time log data, pyro images, time-series sensor data.
  3. Operational excellence: Industrial Artificial Intelligence improved sustainability by reducing emissions. AI increased asset utilization and optimized field services for preventive/ predictive maintenance.


Artificial Intelligence for cement plants can predict, prescribe, and autonomously control cement manufacturing in a supervised-steer mode by utilizing streams of data from thousands of sensors. Industrial AI hopes to improve further by adapting and self-learning from more data and new sources.


In discussion – AI in the cement sector | David Perilli, Global Cement Magazine

Click here to connect with us

Spread the word

Related posts