<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=323483448267600&amp;ev=PageView&amp;noscript=1">

Data For Sustainability

Episode 16

Data For Sustainability

espresso icon Espresso 4.0 by


Sustainability is the hottest topic this time of year or the past few years. I wanted to ask, in your opinion, what is in particular in the manufacturing industry? What is the benefit of data when it comes to tackling some sustainability problems? 

Achieving Sustainability

Sustainability is a big word that is often used these days. You have to know what is behind sustainability. So for industry, what did it mean? Sustainability is producing inhibiting less CO2 in one sense. You can use the same materials input and optimize the output, or you can consume less of your consumables like fuel, water, and gas and still have the same production targets.

Nothing to kick us into more sustainable initiatives than rising prices of energy. Or sometimes it is just water. For example, water in some places is expensive and difficult to find in some remote places. Sometimes optimizing your water consumption is a road to more sustainable production.

For example, two weeks ago, I was in a mining area in Canada, and water is not a problem there because it rains a lot. 

They have these huge pits and blast water on the rocks to get them processed and produce specific products. So much water was infiltrating the ground and pushed outside the pit's walls at high pressure. 

So you have waterfalls coming from the walls of the pit, and they can pump this water and recycle it to use it in production. Also, they have so much of it that they can vaporize the tracks where the trucks are on so they don't produce too much dust for the neighborhood. 

They can afford to “waste” water like that, but you need to optimize the water consumption in some places where there is less. 

The Power of Data in Sustainability

Data can help us in the usage, or rather in reducing the number of resources we use in production, making it more efficient. And the big resource, of course, is energy which comes in different forms. How can they help us do that?

Usually, production processes are already quite well optimized because engineers know what to do to have a production that is controlled, stable, and produce the best possible outputs. So everything is quite well configured because they can see things and have their empirical way of modeling everything that's happening with their systems, physical equations, and thermodynamics.

What can happen is that you have some subtle variation in the process that may not be captured by the human eye or by classical mathematical modeling. 

Those can be just subtle variations in different things not matching simultaneously. But the overall effect can impact the production material that is used.

When you say classical mathematical modeling, do you mean expert systems?

Yeah, expert systems or even deterministic sets of equations. We will be able to do a bit more, maybe with machine learning techniques or a probabilistic approach, to group all these small variations that cannot be captured by the human eye or these deterministic systems and interpret these small variations and evaluate the effect on the production output. 

Energy Optimization

So, for the energy optimization process, we look at some periods where we achieve good production with reduced energy consumption. We will use all the context data we can have from the machines and the production lines or some other type of data like the ambient earth departure and the humidity.

All these can have an impact, and we can, based on machine learning techniques, extract these ranges of value use, even if they are subtle and difficult to catch, and potentially use them to find out that in this specific condition, you can use less energy and still achieve the same. 

What you mentioned is that the manufacturing industry has optimized things quite well up to a certain point, and there is a really small margin left that you can optimize. 

But given that the steel industry is the third or the fourth of the top five biggest energy users in the world, half a percentage of efficiency can be hidden in their data. It could be looked at through machine learning and result in huge benefits both for the environment as well as reduction of cost. 



Sustainability in the Steel Industry

Having mentioned the steel industry, how can these solutions be implemented? Where exactly within the steel manufacturing processes? And how would some go about starting a project like that? What would they need? 

The first thing is to identify where we can have such savings. So everything that consumes gas or fuel can be targeted, like heating and furnace sections. All these things consume a lot of heat. 

For all of that, you can find these different sources. You can check the data that you have in these systems. Do you have data about all the equipment itself - temperature measures, pressures, humidity, air, environment, vibrations, gas compositions, etc.

Also, consider the property of the material that is processed. Based on the chemical composition, you can have different treatments or different reactions to heating. And of course, you have to monitor the output of the production because if you produce some products that reduce energy but the quality of a role is not good, you don't want to learn from it.  

You might have to rework it. Then you need to spend more energy, and so on, or discard it, which is even worse. Sometimes it's even a full reprocess. They have to scrub the material, which they can recycle, but it means going into the blaster and redoing the full loop, including all the steps that consume a lot of energy.

When you're looking into cases in data where you can save on energy, you must maintain adequate quality of the output at all times. Otherwise, it makes no sense. 

First Step Towards Sustainability

How would they launch a project like that? Where do they start? Is there a pilot or feasibility assessment? Is it a case-by-case scenario? Are there prepackaged solutions?

We can do some feasibility assessment if we have a data set with the three features we mentioned before - the process, material, and output data. 

We can find some quick insights from the data with some prepackage algorithms. So we have some out-of-the-box tools that can more or less apply to every case, but we'll give you a view of whether it is feasible and an idea of how much you can win if you apply this outcome. 

And then, if it's successful, we can go into even more details with a more customized approach to be sure that we get every last it's not present, but tens of percent that we can find in this optimization.