IEEE Future Directions
Related Articles
a) Design of raw material
Humankind discovered the first alloy long, long time ago. Historians place the date somewhere between 5,000 and 6,000 years ago. In the current Iran/India area our ancestors were using copper and discovered that smelting certain copper ore, like Olivenite and Tennantite resulted in a stronger metal. Some went further and discovered the smelting copper adding a pinch of arsenic led to the same result. Indeed, those ores contained arsenic mixed with copper. That was, based on historical records, the first alloy: bronze (arsenical bronze). Over the centuries it was discovered that mixing copper during the smelting with other substances led to even better result and didn’t provoke poisoning of the smelters (you don’t want to breath/ingest arsenic). Best results were achieved by mixing tin, in a percentage around 10 to 12.5%. It took 2 millennia to get the right recipe for bronze and it is still good today (for casting bells).
Over the centuries humankind learned to create several other alloys, the most important one, that fuelled the Industrial Revolution, was “steel”. The British found the good recipe for steel and were able to guard the secret for a few decades, then France and Germany (this latter was able to “invent” the recipe by themselves, the French found it more convenient to use spies and attract knowledge …).
There are a basically unlimited number of alloys, just waiting to be discovered. The problem is that there are “potentially” so many. Researchers are so much interested in alloys because each one has specific characteristics (see in the chart on the side differences for different types of steel).
The discovery has accelerated in the last two centuries and wee have now thousands of alloys fitting different needs (racing cars would not perform as they do without the progress in the alloy department, leading to lighter material yet with much greater strength).
The quest for new alloys is not over, quite the contrary. In this Megatrend new ways of exploring the “virtual” world of alloys play a major role.
The word “virtual” is used for a reason. We know that an alloy is the combination of different materials and this combination can go down to atomic level, meaning that in principle we can create two different alloys by composing them with the same percentage of atoms but placing the atoms in different ways (this is the case of the chalk you used to write on a blackboard and seashells: they are both composed of calcium carbonate but their consistency is quite different!). This can give you an idea of the basically unlimited number of possibilities. Also, with a bunch of elements in the periodic table, we have again basically unlimited combination of them. Proceeding by blind experiments like our ancestors did is time, and resource, consuming. Hence the new approach: let’s build these alloys in the cyberspace (>>> virtual).
Machine learning algorithms are being applied to identify possible alloys that would deliver specific characteristics. An example is Intelligens, a US based company that has developed AI software that can help companies dramatically reduce time and investment in the search of new alloys (they claim a compression of 15 years in a month with saving up to 10 million $ in the quest of a specific alloy). An example of composition at atomic level is given in the first figure and further details can be found here.
The ultimate goal is to use additive manufacturing to create the kind of material that is needed in terms of performance and characteristics at the same time as one is manufacturing the product (or a component). Ideally, rather than looking for a material that would be fitting the requirements and then finding ways of using it in an industrial manufacturing process, a designer will specify the required characteristics and an AI based software will create the material in the cyberspace and then through additive manufacturing the desired object will be created, directly driven from the cyberspace. This would support, as this Megatrend is predicting, an on-demand production.
I do not think that this will become commonplace by the end of this decade but several “pieces of the puzzle” will start to become available. Economy of scale is a big obstacle on the way. We might be seeing this approach for very specialised applications with stringent constraints where price is not the point.
