Klabin Improves Process Quality
with an Innovative Data Analytics Project
Klabin is Brazil’s largest paper manufacturer and exporter and the country’s leading producer of papers and paperboard for packaging, industrial bags and corrugated board packaging. It is also the only Brazilian company to simultaneously supply hardwood pulp (eucalyptus), softwood pulp (pine) and fluff pulp to the market. Klabin has 23 industrial plants in Brazil and 1 in Argentina.
They sought to improve process efficiency and productivity across all of its 24 units by collecting and analyzing process data. However, collecting and processing the data in a meaningful way proved problematic. Challenges included the sheer volume of information, which was sometimes incomplete or fragmented, and the cost to obtain it. In some plants, it was impractical to collect data.
The company hired us to develop a bespoke system to increase productivity and raise the bar for analyzing process data at one of its units. The goal was that this innovative and customized solution would be applied in all 10 industrial units of the pulp and paper business.
The chosen unit was the oldest of the group, located in Telêmaco Borba, Brazil, founded in 1946. Its processes and products had constantly evolved, making it more complex than other plants and demands. Therefore, Klabin needed for a customized platform, as it did not find a market solution that fully met its specific needs.
Using functionalities such as Process Analytics and the Quality Modules developed by Radix, Klabin now has a more agile tool for decision making and process quality control, with machine learning support. Integration with all Klabin's process data sources is a fundamental point in the project.
This innovative project in the pulp and paper industry was developed in a mixed architecture on-premise and the cloud, taking into account the specific needs of availability and connectivity, with interfaces with other systems and features linked to the concepts of Industry 4.0, as virtual sensors and machine learning, consolidating knowledge of pulp and paper production processes in a single system.
"Radix helped Klabin achieve tangible gains in efficiency and productivity. The data analytics environment created opportunities for relevant improvements, and enabled the early identification of root causes of problems that could impact product quality. In this way, it was possible to optimize processes and improve logistics and process stability,” – said João Kudo, Radix’s project manager for the Klabin project.
By crossing-referencing data from multiple sources, this new system can identify problems or gaps, find solutions, and make recommendations to the engineers in charge. The most significant gain has been greater agility in collecting, organizing, and analyzing data with the various tools available in the system.
Commented Julimar Bonicenha, manager of the Klabin project: “We commonly find in the market today multiple initiatives aimed at cloud computing that are developed for a scenario and limited study. Our challenge at Klabin was a little different. We created a discovery and study solution. With it, our engineers will be able to analyze the entire universe of process data. The solution provides insights to discover problems, and helps us focus on the process and quality of our products.
“From this continuous discovery, we transfer knowledge to the solution, which is capable of monitoring the process 24/7 and making operational recommendations when necessary for our operators.
“Even with all the variables involved in the production process, such as the temperature and pressure of hundreds of tanks, and thousands of instruments that monitor the operating conditions of various equipment, the access to data has become more intuitive and simple to use for operators and engineers. The machine learning module, for example, promotes an even more advanced treatment, generating alerts based on control limits and presenting corrective action.
“Using a Digital Twin, a virtual sensor compares the actual value with the value of a standard model, guiding the operator to operate on the model without standard deviations. “