Local insight, digital precision
Digitalisation has been a movement in cement production for at least a decade. Using the terabytes of data generated annually by a cement plant, the evolution of acting on real-time information began in earnest. In 2022 Titan America’s Pennsuco plant in Medley, Florida, moved beyond its initial successes with artificial intelligence (AI) and machine learning (ML).
AI-induced innovations through CemAI at Pennsuco optimised operations in two important areas:
1. Upgrade of the existing maintenance and process control AI system
A new set of over 100 high resolution sensors were installed to measure vibrations, temperature and their deviations. This enabled the plant to use sensors and signals to map a digital model for normalised plant operations, allowing for the AI systems to monitor real-time deviations and as a result, to reduce idling and restarting time of major equipment as well as provide significant thermal and electrical energy savings.
Based on previously collected data, AI-based process control solutions were also modified for finish mill process control systems that allowed a higher throughput at each mill and decreased electrical energy consumption.
2. AI solutions at the kiln contributed significantly to stability improvements with increased AF use
Considerable portions of the alternative fuels (AF) are produced in Pennsuco’s own ‘processed engineered fuel’ (PEF) facility. This state-of-the-art PEF facility can receive a wide array of waste streams and was already integrated with advanced technical controls that ensure consistent fuel properties. However, in 2022, it was engaged in the maintenance digitalisation process to improve the system’s reliability and efficiency.
Through these projects and building on years of substantial digitalisation efforts, Pennsuco achieved a new level of optimisation, combining the process and maintenance AI systems to achieve increased overall equipment effectiveness (OEE), better process stability with higher AF rates and lower energy consumption. These projects contributed to decreasing specific electrical energy consumption by six per cent and almost doubled the AF consumption while delivering a seven per cent production increase through the transition from Type I/II to Type IL cement, breaking an all-time cement production record.
Predictive maintenance benefit
One potential benefit of employing an AI system in cement production is the successful prediction of potential failures of the major equipment that could result in expensive stoppages and loss of energy due to preheating/reheating of the equipment or reduction of idle time. CemAI designed the AI modules Pennsuco used in conjunction with its maintenance and production teams. It combines the power of AI and local expertise onsite for validation and proactive corrective actions.
The example of the vertical roller mill (VRM) separator at the Pennsuco plant provides a clear illustration (see Figure 1) of how predictive maintenance can prevent costly equipment failures.
Optimising equipment performance
Pennsuco has now adopted the use of AI’s advanced algorithms, real-time data monitoring and ML techniques to optimise the grinding process. Its AI system continuously analyses several data values to dynamically adjust the grinding controls to increase throughput and reduce energy consumption while maintaining the quality of Titan Florida products.
enabling AI systems to detect and alert potential failure at an early stage
Hidden correlations are revealed when the power of AI can work on all data changes simultaneously. The AI system optimises production by looking at such parameters as, for example, the feed, data from the fans, and the final product quality. It uses a repeating cycle that reads variables, predicts future values, optimises for throughput and efficient energy use, and, ultimately, develops new setpoints.
“AI-driven innovation arms the engineers with precise data and shows new interactions and correlations that they can use to effect real-time adjustments and control,” says Carolina Corzo Ayala, process engineer at Pennsuco. “The technology gets ‘smarter’ with each run...and so do we.”
Pennsuco’s efforts were assisted using an AI-based process-optimising software tool. In their process department, the team is responsible for setting constraints, limits and “collars” to the variables as well as tuning parameter weights. By tuning the weights, the system is aligned to prioritise its decision-making process. For the manipulated variables, limits are defined that should not be crossed. The slope of the steps that can be acted upon is always constrained to safe zones.
Optimiser software works on all data changes simultaneously. The system reads variable values and predicts future values and draws optimised set points to ensure consistent throughput and minimise energy use.
“There is no substitution for human engineering control over the ML/AI optimising processes,” says Pennsuco’s CE director of maintenance engineering, Stojanche Milevski.
Improving Pennsuco’s AF use while maintaining stable kiln conditions
Pennsuco engineers took the opportunity of its completed AF facility to study and adapt the integration of the new fuel streams. In 2022 innovation, design and application merged in Pennsuco’s ambitious AF programme. The on-site PEF facility, completed a year prior, offers world-class technology for engineering fuels, enabling the receipt and processing of a variety of waste streams. Paper, plastics, construction and demolition waste and other waste streams, are analysed, treated and subjected to technical controls that ensure consistent fuel properties.
Control of AF substitution
While the processing of the AF is an essential element, it is only part of a successful fuel substitution. Pennsuco has integrated its AI system into the process to achieve better throughput and avoid adverse impacts from using varied AF streams.
The AI system was trained for a variety of fuels (over five different fuels) and notes the type of fuel selected and the calorific value of each.
The systems also simultaneously manage the fundamental kiln parameters, such as the fuel ratio, burning zone temperature, kiln load and quality of the product as well as other parameters important for the stability of the kiln operation.
Fortified with real-time data from within the kiln, AI and the operations team use predictions of these variables to enable proactive action on the major process control variables. As a result of these improvements, Pennsuco was able to consume more than 36,000t of solid alternative waste in the case study year.
AI solutions enabled the plant to operate under more stable kiln conditions, directly improving the AF substitution rate and optimising the specific thermal energy consumption, despite the variability of AF quality. In this way, the beneficial impact on the environment is two-fold – not only in terms of AF use but also in terms of overall energy reduction.
What can others in the industry learn from this project?
Working at a level of Digitalisation 4.0, Pennsuco amassed a large volume of operational data, applied AI/ML to develop correlations and interrelationships revealed by the program’s analytics and effected changes to procedures based on the new understanding obtained.
After start-up success in prescriptive maintenance, the plant added 100 new, wireless sensors to the system in 2022. The determination of data sensor locations and the significance of new data, and any new correlations discovered, requires the local expert engineer to upscale their assessments and innovative thinking. Pennsuco has learned that the power of AI is best harnessed when knowledgeable, local teams lead the design function.
The new sensors that were placed on selected equipment by the production engineers increased visibility in equipment previously outside the digital model. Upon incorporating these new sensors, the plant saw benefits in the short term. Vibrational characteristics and “mission-critical” status of equipment were the principal selection criteria.
The AI prescriptive maintenance system is based on the known curve for failure development and aids in detecting the failures at the stage when there may be only data faults, recognised by a certain set of process or condition monitoring parameters within the AI-based system.
The AI system applied its algorithms to the data obtained; engineering teams design the sources for data collection and interpret the data obtained. Predictive maintenance coupled with prescriptive maintenance practices based on AI/ML resulted in energy savings.
Another “take-away” is the value of local engineering oversight to AI optimisation processing. Pennsuco engineers devised specific “collars” to changes that the optimisation software could implement to affect the grinding function, for example.
The AI system follows a four-step cycle that repeats every 30s, and the program will:
1. read the current value of the variables
2. predict future value of the variables based on neural network and empirical models
3. optimise by maximising throughput and minimising energy consumption
4. write new optimised set points on manipulated variables, based on its predictions or to correct a constraint.
The use of “collars” set by the knowledgeable local engineering teams and the “scale” that develops from optimising software continuous feedback is truly a potent combination of manufacturing knowledge and know-how.
Conclusion
The digitalisation revolution has brought unprecedented insights into the cement manufacturing process, transforming how data is captured and utilised. With CemAI’s advanced AI and ML technology, real-time operational data – such as temperature, vibration, and lubrication status – can now be analysed and acted upon more effectively. This allows manufacturers such as Pennsuco to make data-driven decisions that enhance efficiency, reduce energy consumption, and increase output.
By leveraging this wealth of information, CemAI’s solutions enable plants to continuously adapt and optimise their processes. “As we obtain more data, which aids a greater understanding of correlations and interconnections of various functions, we can safely improvise with processes like grinding and managing alternative fuels,” says Zaklina Stamboliska, vice president of cement manufacturing.
The integration of CemAI’s AI systems, alongside the expertise of local teams, is driving a new era of intelligent, sustainable production. As demonstrated at Pennsuco, this technological evolution, supported by CemAI’s digitalisation tools, is not only improving operational control but also shaping the future of the cement industry.
This article was first published in the November 2024 issue of International Cement Review.
