In this case study, we present and discuss HABER's approach and AI's role in addressing one of India's leading pulp manufacturer's concerns around maintaining the final tower pulp brightness within the specific target range, reducing costs incurred due to excess consumption of bleach chemicals.
Variability in the quality of wood chips and bleach chemicals, anomalies in measuring equipment such as brightness sensors, and variations in the process parameters resulted in over-utilization or under-utilization of bleach chemicals, leading to deviations (+/- 5% ISO brightness points) in the final tower brightness. The mill uses chemical and mechanical pulping processes. Pulping process involves four stages namely pre-Bleaching, refining, bleaching and washing. In Pre-Bleaching stage the chips are preheated with steam. In refining stage, the pulp is refined to attain certain degree of freeness. In bleaching stage chemicals such as peroxide, caustic, and chelating agent are used to achieve desired bleaching levels. In washing stage, also the final stage of the pulping process, the unused chemicals are recovered before transferring to the final pulp storage towers.
HABER’s approach included four steps in achieving the goal.
The first step in our approach was to identify all process parameters affecting the tower brightness and collect the data for all critical parameters measured through both online sensors and offline measurements in the lab.
At the end of this step, with HABER's eLIXA platform, we were able to
The second step in our approach was to analyze and make necessary interpretations from the data. Using state-of-the-art analytical tools such as Topological Data Analysis and advanced statistical tools, we were able to identify different clusters in the data, along with many other hidden patterns. These were then used to select the most influential parameters and feature generation. This brought the dimension of the data down to ~35 parameters.
Some of most influential parameters found to impact the brightness are
The third step of our approach was to build a predictive model using advanced machine learning and deep learning tools. Since the brightness is measured at different locations across the process, we have designed different AI models for different stages. Our AI is designed to control the output brightness within certain ranges at each stage of bleaching process. A strict control at each stage allows us to maintain the final output quite steadily, thus reducing the variation in the final brightness, while also allowing for optimization of the chemical dosages. With the integration of HABER's eLIXA platform with distributed control system (DCS), the set points for chemical dosage at different stages that are computed on real time basis from the models are fed to DCS and then appropriate signals are sent to the final control elements or dosing pumps.
And the final step of our approach is the digitization of the data. HABER's interactive data visualization and analytical dashboards with user focused objectives provides platform to track, analyze and display key performance indicator (KPI) metrics & real-time insights. Behind the scenes, HABER's dashboard connects multiple sources of data and on surface displays tables, line charts, bar charts along with functions such as drilldown options. From the insights section (that incorporates AI tools), the operating crew identify and correct negative trends that may be occurring. HABER's analytical dashboard also became an integral part of the manufacturer reporting process for plant performance summary, equipment performance highlights, and chemical consumption reports.
This is a classic example of industry 4.0 application where interconnectivity, comprehensive evaluation of data from many different sources starting from production equipment to lab measurements, and artificial intelligence become standard to support real-time decision making.