Lessons Learned from Global Deployment
In building and deploying a real-time infrastructure from their many sites, MaterialsCo recommends the following six best practices:
- 1. Capture needs and proposals from the operations' network to build the initial list of projects in a collaborative way.
- 2. Confirm that all proposed initiatives align to company objectives and business impacts.
- 3. Implement extensive testing programs because testing is always productive and never a waste of time.
- 4. Build a showroom for testing and demonstration purposes, with the least amount of resources.
- 5. Document value obtained from deploying each initiative.
- 6. Continuous improvement is an ongoing cycle: Keep moving forward.
Moving to the Next Level—Artificial Intelligence and Closed-Loop Control
MaterialsCo's latest initiative builds on having used their operating data for descriptive analytics and understanding why things have happened. They are now leveraging their data to develop predictive and prescriptive analytics.
Predictive analytics uses statistical models and calculations to understand and forecast what could happen in the future, such as when specific assets (equipment) will fail. They are also developing prescriptive analytics, in the areas of
- • Operations efficiency, forecasting incidents, and downtimes;
- • Energy optimization;
- • Quality assurance;
- • Safety; and
- • Alternative fuels substitution.
MaterialsCo is now implementing artificial intelligence (Al)-enabled autonomous control for certain cement plant operations using PI System data with AI tools developed by a software analytics company. Their goals are increased efficiencies leading to lower costs, reduced fuel and energy consumption, better quality, and improved decision-making. The initiative is a phased approach consisting of the following:
- • Predict. Real-time forecasts improve understanding.
- • Prescribe only. Prescriptive instructions are validated by plant operators before updating kiln set points. The Autosteer capability is turned off in this mode.
- • Autonomous control. The Autosteer mode is turned on, with supervised, controlled autopilot operation, integrated with plant control systems.
MaterialsCo has currently achieved the following milestones with this initiative:
Phase I. Forecast predictions in real time. They successfully predicted clinker and air temperatures 15 minutes in advance.
Phase II. Real-time prescriptive recommendations. They relayed recommendations to kiln operators, with model improvements to make recommendations reasonable and actionable.
Phase III. Autosteer operation of the kiln's cooler section. They successfully ran the kiln cooler autonomously, with higher exit air temperatures when Autosteer was engaged.
Phase IV. Multi-site deployment, which is yet to be implemented. In this phase, they will replicate the process using the PI System infrastructure and identify best practices to improve machine learning models.