15 May Autopiloting the cement process – Cement World May 2024
Fayez Boughosn and Ighnatios Maatouk, ES Processing, explore the benefits of autopiloting the cement manufacturing process.
The cement industry is being reshaped through various challenges such as climate engagement initiatives,digitalization, energy costs and market volatility.
ES Processing developed its first APC-AI CMO as a ball mill autopilot, introducing artificial intelligence (AI) and advanced process control as new levers for cement manufacturing, continuously optimising the mill grinding process. The company’s CMO (cement mill optimiser for ball mills), VMO (vertical mill optimiser) and KPO (kiln process optimiser) solutions are field proven and have consistently evolved with the latest technology.
Autopilot solutions
The CMO, VMO and KPO solutions represent a complex combination of AI, machine learning and advanced process control, able to predict product quality every 30 seconds instead of performing routine sampling every hour or more. They then perform the proper actions on the kiln and mills’ manipulated variables every 30 seconds. Predictions are continuously adapted to the new process conditions such as quality of throughput material, feed rates, rejects, pressure, etc.
These solutions help transform the cement manufacturing process into proactive, autopiloted operations, embedding continuous accurate predicted product quality into a continuous decisions matrix of process optimisation.
Process optimisers
The processes in cement manufacturing, especially clinkerisation and grinding, are highly nonlinear and unstable, with directly impact product quality, plant performance, and operational costs. Therefore, the main target behind ES Processing’s advanced optimisers (CMO/VMO/KPO) is to enhance the production of uniform clinker/cement at the targeted quality. These solutions optimise the product quality, significantly reducing variations, whilst boosting the output and stabilising the overall process.
The CMO, VMO, and KPO solutions act as autopilot systems for the cement kilns and mills, enhancing the clinker reactivity and the cement fineness whilst ultimately increasing overall production and reducing the specific energy consumption. They also continuously determine the optimum setpoints for the mill/kiln manipulated variables in order to establish a seamless operation while ensuring product quality and optimum performance.
Each solution’s main strategy is to provide inferred quality measurements, at a high frequency (every 30 seconds), to be used as a decision support system, helping to implement proper adjustments to selected manipulated variables. This speeds up operations with increased stability by reducing the traditional dead time caused by the lack of continuous quality analysis.
These solutions are composed of a complex combination of:
- Soft sensors: These are very sophisticated but stable models formed by combining multiple data-based algorithms adopted from machine learning and based on linear and non-linear identification techniques, PLS and genetic algorithms. They determine the best correlation between different process parameters and product quality results and thus are able to predict very accurately the main product quality indicators (blaine, residue, C3S, free lime) every 30 seconds.
- MPC: This is a highly complex multivariable model based on transfer functions built according to impulse tests results performed on each equipment. It is able to handle complex plant dynamics including long-dead times and non-minimum phase behaviour, constraint handling, hierarchical and weighted optimisation and predictive control. Thus, it can adjust the process manipulated variables every 30 seconds.
With this combination, the plant is consistently pushed to achieve its operational targets, driven by the continuous prediction of quality product and process reactions, along with the optimal adjustments of operating conditions. It is like having a dozen of the best highly-skilled operators continuously monitoring all process parameters, including online product analysis, and working simultaneously together to adjust the process parameters every 30 seconds to optimise the entire process.
Product quality prediction
The product quality has always been the weakest part of the chain in cement manufacturing due to many factors including: process time delay, sampling challenges, analysis accuracy as well as operation techniques.
Through understanding that the continuous monitoring of product quality helps to optimise the processes, ES Processing introduced its soft sensors to predict, every 30 seconds, the clinker/cement quality. Numerical data-based algorithms, adopted from machine learning and AI, build the core of the soft sensors. Adapting engineering correlations between the plant historical process and quality data, the soft sensors are used to infer cement fineness (blaine and/or residue) and clinker quality (C3S and/or free lime) from a defined set of inputs that show a strong correlation with the outputs.
Eventually, these soft sensors act as real-time product quality measurement devices, providing continuous feedback to the control systems even in the absence of laboratory data. Once laboratory measurements become available, the soft sensors use the results to dynamically improve and tune their predictions.
A chess strategy for advanced process control
Under normal operation, the CCR operator checks the product quality analysis and, based on other process indicators, selects a combination of actions on the kiln/mill. Technically, a skilled operator would have to decide and make calculated adjustments by using their experience to understand the possible process reactions. For improved performance and stable operation, a computational tool is required to exploit the current process indicators and past conditions of the plant in order to predict precisely the future plant behaviour and calculate a combination of control moves.
The MPC module of the CMO/VMO/KPO solutions, built based on advanced numerical algorithms derived from the plant data, applies the famous chess-game strategy, consisting of taking continuous and simultaneous actions (every 30 seconds) on the different manipulated variables of the mill/kiln while always predicting the process reaction for the next 120 moves (one hour of operation).
Eventually, the MPC actively controls and stabilises the mill/kiln in real time, aiming to produce a uniform product at the best targeted quality along with an optimised performance resulting in increased volumes at lower costs.
About the authors
Fayez Boughosn is the founder and CEO of ES Processing. Fayez started his career as an Automation & Process Control Engineer for Holcim, where he worked with Siemens to develop one of the most advanced cement production lines of the time. Moving on as a Siemens Solution Partner, he led numerous challenging process control projects worldwide. Coupling his passion to innovate with his cement industry experience, Fayez founded ES Processing in 2007.
Ighnatios Maatouk is the Process expert & APC Projects Manager at ES Processing, managing the most challenging projects of advanced process control and AI solutions for cement plants’ optimisation. Ighnatios started his career as a site engineer for power plants and afterwards served in engineering managerial roles in the cement industry for Holcim as Process Performance Manager. Ighnatios’ ambitions for bringing innovation and AI solutions to the cement Industry led him naturally to join ES Processing in 2018.