6/2021

Along with decarbonation and decentralisation, digitalisation presents one of the major challenges for the cement industry. But it also presents a major chance. The requirements for the cement process arising from the industry’s path towards carbon-neutral production, such as increasing rates of alternative raw materials and fuels or the operation of carbon capture plants are complex. Data-driven solutions and artificial intelligence can help to control and optimise the production process. But first we need to understand and control the data required for this.

Digital tools, such as software for process simulation and the computation of fluid dynamic and computer-aided design and planning tools, are an important part of today’s cement production. Process control expert systems based on fuzzy logic or case-based reasoning have already been implemented in the early ‘90s, and even machine learning models are not completely new since these have been part of model-predictive control algorithms for over a decade. But digitalisation has not only introduced a new generation of smart sensors and data-driven solutions - it has  also changed the way data is handled in the industry.

In control of the process

Smart (soft) sensors can virtually ‘measure’ product fineness or clinker and cement quality without taking any samples. Plant equipment, for example high efficiency SNCR units, is locally controlled and dynamically optimised by machine learning algorithms, and AI applications have also found their way into the control and optimisation of the pyro process and grinding plants. By taking into account more process data than an operator can assess at any one time, these applications can directly react to slight fluctuations, for example caused by properties of alternative fuels/raw materials (Fig. 1). There are also applications known in other process industries where AI solutions can even outperform operators by departing from common methods of process operation.
 

Figure 1: Central Control Room (CCR) operators at work

These improvements can only however be realised on the basis of a process and equipment in faultless condition. Control systems can help stabilise and optimise process performance, but they cannot compensate for severe mechanical problems like uneven material flows, defects in the feeding systems or a blocked outlet diaphragm. Before the introduction of new control solutions, classic process investigations are recommended. Not only the quality of the equipment, but also the quality of its data is a key factor for the successful introduction of data-driven solutions for process control.

In control of data

The quality of datasets refers to a variety of properties. First of all, measured values need to be correct. Sensors need to be calIbrated and correctly parameterised. All data measured needs to be assigned to universal timestamps. Even perfectly correct data will however show sudden jumps, stops or unusual values. To be able to explain these deviations, information on disturbances or changes such as the replacement of parts or the introduction of new materials needs to be documented. This is time consuming and may require new documentation strategies, but understanding process data and building up structures for its efficient use are important preconditions for process optimisation and control.

Digitalisation in cement plants today is often driven by stand-alone solutions bringing only a limited amount of data-driven services to the plant at a time. The vision of the digitalised cement plant of the future incorporates the intelligent and standardised management of all data. This will involve unified interfaces and denomination of data, but also concepts for storing data with additional information which allows better navigation and interpretation of datasets.

Training and Acceptance

Despite all digital tools, CCR operators are and will still be key persons to ensure a stable and efficient production process. They continuously supervise the process and are ready to intervene in critical situations when experts systems are overstrained. Digital tools and AI applications are meant to make their life easier and help them focus on optimisation rather than simple control. But in many cases acceptance for new solutions is still low. Sometimes expert systems fail at the start because they are not yet fully adapted to the specific process. Sometimes smart measurement systems deliver wrong values and need to be recalibrated. It is easy to say “I told you so” and go back to the conventional way of doing things, but at this point operators, process engineers and data scientist need to work together, fine-tune and further develop the digital tools. Establishing such a mindset requires time, transparency and above all training. It is essential that all staff involved gain insight into the methods behind digitalisation projects, their added value and the required preconditions. Training often needs to start with basic vocabulary. Not everyone is aware of the “IoT, AI and APIs”:

Besides training on the basics of digitalisation, the digitalisation of training itself is gaining more and more importance. E-learning is already a widespread approach towards the qualification of operators. Latest developments include the use of Augmented Reality (AR) or Virtual Reality (VR) to provide even more realistic training experiences. For process control, especially simulator-based scenario training on critical situations will be crucial in a future with an increasing degree of automation (Fig. 2).

Figure 2: In simulator training virtual plants are connected to real control environments.

Preparing cements plant for digitalisation is a challenge. Preconditions regarding the equipment, data quality and also the staff involved must be met. In return, digital tools will help the cement industry to take on the even bigger challenges of carbon neutrality and resource-efficient production.