NTWIST regularly attends multiple conventions around August to meet industry professionals, discuss their pains and solutions, and share our vision. This year, we started our conference tour by attending the 19th IFAC Symposium held in Montréal on August 15-17 . The event covered a lot of topics of interest, such as Control, Optimization and Automation in Mining, Mineral and Metal Processing. We’d like to share some of our notes.
1. Real time optimization for better control performance
Maintaining a good process economy through unrealized process inefficiencies/practices remains crucial today to sustain growth in the manufacturing industry. The good news is that it does not require additional investment to build new infrastructure.
Eduardo Nunez highlighted how APC (Advanced Process Control) could be leveraged with analytics, big data, and AI/ML to maximize the benefits through digital transformation. He also addressed the questions like common mistakes, expectations, and dos and don’ts inspired by ten real-life operating-floor implementations.
2. Advanced Process Control Is Our Future
The field of Industrial Process control has come a long way in the last 80 years, from simple PID controllers to model-based multivariate control. The complexity of the control has risen dramatically to drive the process towards optima. However, Sigurd Skogestad highlights that sophisticated control architecture may not always be necessary to handle complex tasks.
Skogestad showcased with examples how the Conventional Advanced Process Control (APC) elements, such as cascade control, split range control, and multiple controllers, can be proposed to enhance the control and handle constraints without non-linear dynamic models and optimization. He illustrated how using APC can simplify the solution in certain cases to avoid the challenges presented by model-based control. In the end, He stressed that in modern process control, both APC and MPC are in a parallel universe and needed in the control engineer’s toolbox.
3. Operation Participation Is The Key Of Future Improvements
Making a timely decision in plant operations is essential for optimal operations. Sherritt Technologies and NTWIST have joined the forces and developed a data-driven soft-sensor using machine learning algorithms to address this problem. Careful multivariate data analysis of 2 years of historical plant data was combined with a customized novel framework with Neural Networks to develop these soft sensors.
The main goal of this development was to eliminate the bottleneck lab measurement that prevents efficient decision-making in plant operations. Maryam Azin demonstrated the real-time performance of these soft sensors along with the results of the sampling campaign to ensure validity. This solution provides a minute-by-minute inference of the KPI that otherwise would require expansive, complex, and time-consuming lab analysis.