Welcome   Guest · ·

AL Circle Blog

Aluminium Industry Trend & Analysis, Technology Review, Event Rundown and Much More …

Aluminium Industry Trend & Analysis, Technology Review, Event Rundown and Much More …

AL Circle

Data science-based AI-driven predictive model: An effective solution for primary aluminium industry

Contributed by:

The primary aluminium industry will require data science-based, innovative artificial intelligence solutions that are integrated with conventional techniques and non-contact evaluation methods.

Reading data to interpret a trend is of huge significance for any process industry, where even a small deviation may result in significant losses for the organisation. Decades ago, statistical process control by plotting charts with reliable data was also an effective mechanism wherein we were able to understand process capability and avoid the complete halt of a machine because of the failure of its components.

In the spectrum of Artificial Intelligence, both Machine learning and Deep learning are techniques that can be used to understand a complex process and the hidden correlation of different process parameters.

While both are algorithms that use data to learn, the key difference is how they process and learn from it. For example: Machine learning models need human intervention to learn from behaviours and data. Deep learning models use neural networks to adjust behaviours and make predictions.

Recently, in manufacturing and processing, people have found it very useful to create a database and analyse the same to point out excellent mathematical correlation which can save billions of dollars.

Wondering how? Let’s take an example of building a Predictive Model for predicting a possible reduction in the life of an Aluminium electrolytic smelting Pot. This issue is one in Aluminium Pot operation where stricter operational guidance is going to be very useful. There are multiple parameters influencing the failure of wall liners operating under severe conditions of approximately 980 degrees Celsius under molten alkali fluoride salts.

While stoichiometric ratios of alkalis and fluorides of different cations are extremely important, it is very much essential to utilise data on temperature fluctuation, non-contact measurement of shell temperature and expansion-contraction characteristics of the same.

In addition to this, other factors such as FEA analysis of cathode Blocks and the possible chemical potential of electrolyte reactions with liners will be essential for gaining insights.

Much of the innovation implementable on the floor will be more complex than what we can anticipate and therefore this exercise will be a long-term project and will give rise to a Digital Twin with effective prescriptive control.

JOIN OUR NEWSLETTER
And get notified everytime we publish a new blog post.

Leave a Reply

Your email address will not be published. Required fields are marked *