Earlier this year, Howden’s Lead Software Engineer, Benoit Dussault, told IM that the company was starting to delve into machine learning as part of the evolution of Ventsim CONTROL, and he recently provided a few more details about the impact this could have on its flagship ventilation optimisation system.
The company’s aim for this project – as with all ongoing ventilation projects – is to optimise the flow of air and the other varying parameters that come with adequately ventilating underground mines.
Ventsim CONTROL has proven in the past to improve the ventilation process, with its highest level solution – level 5 – offering a ventilation on demand (VoD) solution bolstered by “optimisation algorithms”.
Dussault said the company is looking to bolster these algorithms with machine learning to help predict and detect certain parameters that influence the way mine ventilation systems work.
“With that, we could detect something that a sensor alone cannot do and analyse data to see things we could not see before,” he explained.
For instance, with Howden’s recently added temperature controllers for Ventsim CONTROL – both for cooling and heating purposes – the system could leverage machine-learning algorithms to predict how long it would take to reach a specific temperature sub point at an area of a mine, optimising the heat and airflow so that the set point is reached at the scheduled time.
This type of process reduces not only the energy consumption associated with ventilation but also the emissions associated with powering the processes.
“There is already optimisation happening with Ventsim CONTROL on a regular basis, but, with the assistance of machine learning and predictive modelling, we can optimise this further,” Dussault said.
As both software developers, engineers and mining practice leads, Howden, a Chart Industries Company, is well positioned to make the most of the industry’s machine-learning advances, according to Dussault.
“In the mining industry, there are a lot of PLC programmers and automation specialists, but very few of these are software developers,” he said. “I think we have much unique expertise to allow us to lead this adoption.”
To facilitate this move, Howden is moving Ventsim CONTROL over to a web-based user interface with BI Dashboards and reporting, making it easier to understand what the data is saying about potential ventilation optimisation advances.
Howden is currently evaluating the customer needs to build a machine-learning prototype that will be tested extensively in-house ahead of deployment at a mine site.
Feeding the algorithm with the right kind of data will be paramount to the project’s success, according to Dussault.
“No matter what you try to do with machine learning, if your data is wrong, your model will be wrong,” he said.
