Increasing productivity in heavy industry processing plants with advanced analytics
A major copper concentrate producer was struggling to run grinding mills at its ideal capacity and to achieve top-decile recovery performance. The challenges were driven by the plant’s complex operations, where thousands of sensors were fed to a control system to solve for non-linear functions and trade-offs. We created a custom-built artificial intelligence model–loaded with three years of operating data–to identify operational tweaks that would boost copper production at consistent quality. As a result, the plant improved throughput by a range of 10 to 15 percent, and copper recovery in a range of two to four percentage points.
A leader in biochemicals aspired to embed digital and analytics in its way of working. Together, we launched a tech-enabled manufacturing transformation, initiated with the deployment of five global centers of excellence and sustained by training the majority of the practitioners. Through a series of agile sprints, we developed advanced analytics models to identify the most important drivers for process variability. Automated data pipelines and interactive dashboards enabled embedding this data-driven decision making into daily operations. As a result, operators were able to reduce process variability and to increase throughput by 10 to 15 percent in bottlenecked sites.
A large North American pulp mill operation was only achieving 60 percent of its target production at sub-optimal quality in final product. We leveraged AI to review the design of existing control strategy, including advanced process control (APC) loops, at key unit operations in the mill. It helped identify the gaps in the control design and operating philosophy. AI also helped in identifying the desired operating ranges—including set points and upper or lower tolerances—for key process variables in the APC in close relationship with subject matter experts. This resulted in a 15 percent increase in throughput and four percent increase in yield.
A leading European refining operations company was aspiring to improve one of its top-performing assets using state-of-the art analytics tools. The asset already had industry leading analytical tools in place such as unit level APCs, kinetic models, and site-level linear program optimizer.
We partnered with a global cloud computing leader to develop custom-built machine learning models and tools to improve prediction of quality specs, process indicators, constraints, and unit performance. These models served to create a global optimizer to minimize quality giveaway while maximizing throughput. It helped achieve a $0.3/bbl margin improvement (equivalent to five percent variable operating margin).