IDEAS Intelligent Control

“It’s a tremendous opportunity for us to deploy a pilot installation,” says Sohail Nazari, ANDRITZ AUTOMATION’s business development manager. The AI training concept was developed by a team of ANDRITZ control engineers who first entered an internal innovation contest judged by ANDRITZ’s executive board. As the winners, they were given funding for a prototype and the opportunity to focus on developing the new technology. The recent win at #DisruptMining will enable them to carry the concept further. “Currently, our artificial intelligence has run only in a simulated environment,” says Nazari. “With GoldCorp’s support, we will be able to take the important step from the laboratory to the plant.” 

About #DisruptMining

Now in its third year, #DisruptMining provides a platform for entrepreneurs to showcase disruptive and exponential technologies that have the potential to accelerate innovation in the mining industry. From approximately 90 submissions, six semi-finalists and three finalists were chosen. The #DisruptMining live finale took place during the annual Prospectors and Developers Association of Canada (PDAC) conference held on March 3. Semi-finalists were invited to display their work at PDAC’s innovation expo, while finalists presented a pitch video and participated in a live Q&A session before an audience including mining executives, investors and financiers, all eager for a glimpse into the future of mining.

ANDRITZ’s Winning AI Innovation

ANDRITZ’s simulation software IDEAS enables ANDRITZ to build a fully dynamic model of a processing plant based on first principles. ANDRITZ uses these digital replicas for various applications, one of which is the creation of a control room Operator Training Simulator (OTS) that is similar to a flight simulator. For over 20 years, ANDRITZ has used the OTS and IDEAS training software to train process operators. Then, a breakthrough.

“We thought, if we’re able to build training simulators so accurate that they can train operators, why not use the same environment to train artificial intelligence?” explains Nazari, adding, “AI is like a baby, it has to learn by making mistakes.” The challenge was creating an environment in which it could do just that.

 To accomplish the task, ANDRITZ used IDEAS to develop a digital twin of an entire mineral processing operation. In this safe virtual environment, the AI can make changes, observe the consequences and update its priorities. Over many iterations, it learns the best operating points, how to start up and shut down, how to process minerals and optimize production, and how to respond to abnormal and emergency situations. Once fully trained, the AI can be integrated with the real-time operation.