What does a cement plant of the future look like? (Part 2)

By | 4 minute read | December 1, 2017

In the first part of this two-part series, we explored what defines a cognitive manufacturing plant.

This blog will look at a specific use case -a cement manufacturer – to get a sense of exactly how this would work.

Cement-making requires precision and little room for error

Cement manufacturing has three key process steps:

1) Limestone from the quarry is crushed and appropriate raw material, like iron oxide, silica oxide, aluminum oxide, etc is mixed and ground to get the raw meal.

2) This raw meal is passed through the cement kiln at very high temperatures to produce clinker.

3) The cement mill grinds the clinkers to an appropriate heat to produce cement.

Each step is critical but the final step is where the magic happens.

Grinding to perfection

In the final step, the cement mill grinds the clinkers using horizontal ball mills. The ball mill has two rotating chambers with ceramic balls that grind the clinkers.  The clinkers are fed from one side.   The mill is heated at various temperatures along the separator.  The temperature evaporates the water in the clinker and initiates various chemical reactions within the chambers. The size of the ball, speed of rotation and duration of grinding will impact how fine the cement is.

While higher speeds may give better grinding, the centrifugal force on the balls can impact the grinding capability beyond a certain speed.  Similarly, the longer the grinding duration, the more fineness it will have, but more energy will be consumed in its creation. This excess energy could exceed quality requirements and lower yield.   Optimization between yield, quality and cost is a fine balancing act which is necessary to maximize return on invested capital (ROIC).

How can a cognitive plant help?

A cognitive cement plant uses advanced cognitive computing to predict variability across key metrics. We base these metrics, including throughput, quality and energy consumption, on  data that we obtain from the processes and machines.  When the predictions are out of range, advanced algorithms are used to prescribe operating parameters that could optimize the production KPIs.

Using these parameters enable the variability in plant performance to be minimized.  Thus a cognitive cement plant is run at optimal performance irrespective of raw material or environmental variants. This can save millions of dollars in energy cost and throughput every year.

IBM Plant Advisor recommends ways to reduce energy costs

Grinding cement requires a great deal of energy.  As the fineness of the cement increases, it needs more energy to grind it.  But every cement quality band has a certain fineness requirement.  In one example mill, we consider the fineness and the energy consumed for eleven months for a cement mill. In the first five months, we found that this plant has been producing cement quality more than what is necessary to produce.  As a result, this plant is consuming at an average 15% more energy than what is necessary.

Later, the plant implemented IBM Production Optimization for this process.  As we covered in part one, this solution is trained with the historic data and the system predicts the fineness and energy for a given situation. From there, it starts recommending set points for two of the manipulated variables; rotating speed of the separator and the flow rate for the clinker feed. The operators accept the recommendation and make changes to the control parameter set points.

The net result; the fineness is under the acceptable band and they reduce energy consumption by 12-15% by implementing the recommendations given. Production Optimization also predicts the future value of these variables and plots them.  The plant manager monitors these KPIs and their predicted values to be proactive.

Cementing the benefits of a cognitive plant

Moving to a cognitive plant can yield up to several million in cost savings per year in energy savings and higher return in terms of increased throughput. In addition to that, the Production Optimization solution can provide the following improvements:

  • Helps to retain expert operator skills and leverage them with less experienced users. This keeps the knowledge within the company.
  • The purpose-built machine learning pipelines enable a quick start and early time to value.
  • Offered as a solution-as-a-service requiring minimum on-premise infrastructure, it reduces capital investment and accelerates return on investment.

Going beyond the cement manufacturing plant

Beyond cement, cognitive manufacturing has potential in many areas, such as in steel production, mining, power plants, pulp and paper production and metal smelting.

McKinsey and Company estimates net saving of $1M per year for large processes and $50-100K per year for midsize processes.  For a chemical company with 150 plants, that could mean $50-500M savings per year.

Where to find additional information:

Ready to take the next step on your cognitive plant journey?

Learn more about IBM’s Production Optimization solutions.