Posted by Daniel Gleason on July 11, 2023
Earlier this year, Howden’s principal software engineer, Benoît Dussault, said: I The company says it’s embarking on machine learning research as part of the evolution of Ventsim CONTROL, and recently went into more detail about the impact this will have on its flagship ventilation optimization system.
The company’s objective with this project, as with all ongoing ventilation projects, is to optimize airflow and various other parameters associated with proper ventilation of underground mines.
At its highest level of solution, Level 5, Ventsim CONTROL has proven in the past to provide a ventilation-on-demand (VoD) solution enhanced by an ‘optimization algorithm’ to improve the ventilation process.
Dusseau said the company is looking at using machine learning to enhance these algorithms to help predict and detect specific parameters that affect the operation of mine ventilation systems. .
“This allows us to detect things that sensors alone can’t, or analyze data to see things we couldn’t see before,” he explained.
For example, with Howden’s recently added temperature controllers for Ventsim CONTROL (both cooling and heating), the system utilizes machine learning algorithms to determine how long it takes to reach a given temperature subpoint in the mine area. Predict and optimize. Adjust heat and air flow to reach set point at scheduled time.
This type of process reduces not only the energy consumption associated with ventilation, but also the emissions associated with powering the process.
“Optimization with Ventsim CONTROL is already routine, but with the help of machine learning and predictive modeling, we can optimize this even further,” Dusso said.
As a software developer, engineer and leader in mining practices, Chart Industries’ Howden is well-positioned to take full advantage of the industry’s advances in machine learning, Dusso said.
“The mining industry has a lot of PLC programmers and automation experts, but very few software developers,” he said. “I think we have a lot of unique expertise that can lead this implementation.”
To facilitate this move, Howden migrated Ventsim CONTROL to a web-based user interface with BI dashboards and reporting capabilities to understand what the data showed about potential ventilation optimization advances. We make it easy.
Howden is currently assessing customer needs to build machine learning prototypes that are extensively tested in-house prior to deployment at the mine site.
According to Dussault, feeding the right kind of data to the algorithm is paramount to the success of the project.
“No matter what you do with machine learning, if the data is wrong, the model is wrong,” he said.
