Mark Yseboodt, Siemens, explains the importance of embracing machine learning and software innovations in order to drastically improve cement plant performance.
Cost reductions, environmental protection, energy efficiency and capital optimisation are among the major challenges that most industries are facing today and the cement industry is no exception. The solution to these, and other burning business challenges, involves embracing the latest technology developments and innovations, particularly in the rapidly expanding software environment. The focus today is on collecting, managing, displaying, and analysing data to such an extent that decision-makers and plant managers are in a position to make knowledgeable decisions regarding how best to operate their plants. In fact, the relevance of knowledge is nothing new, since Sir Francis Bacon recognised the significance of knowledge as early as the 16th Century with his famous quote “knowledge is power”. In the 21st Century, data is knowledge. The use of artificial intelligence (AI) and machine learning (ML) in industrial applications takes data usage and analytics to a new level by allowing computer systems to perform tasks which normally require human intelligence and intervention.
In the July issue of World Cement, Siemens showed a first glimpse of the company’s work with artificial intelligence and machine learning in the cement industry. The article discussed the introduction of two new modules for Siemens’ digitisation platform Sicement Operations, namely ML for Smart Anomaly Detection and AI to predict kiln behaviour.
Developments in this domain are now moving at lightning speed and with only a few months having passed since the publication of the article, the solutions are maturing.
AI assisted auto pilot for the kiln 2.0
In the first implementation of the AI solution for a kiln, the system looked for patterns in the historical data of relevant kiln values. With these patterns as basic information, the AI module was able to predict the values of all relevant variables for the next 30 minutes. These values allowed the operator to optimise the setpoints to ensure that the kiln was running optimally, while at the same time avoiding shutdowns due to out-of-range values. The solution was commissioned with six months of historical data and once the system was fully up and running, the AI module continued to collect information to further train the system. An important part of the training is the continuous comparison between predicted values and actual values. Figure 2 shows this comparison – the green line is the predicted value, and the red line is the measured value, 30 minutes later. The results showed that for most of the predictions, accuracies of 85% or higher were achieved.
Found this article interesting? You can read the full version in the December issue of World Cement.
To register for a free trial, click here!
Read the article online at: https://www.worldcement.com/special-reports/10122021/data-is-power/