Some years ago, I wrote this column focusing on the possible use of machine learning and artificial intelligence for shot peening set up and tuning. More in detail, if I remember well, I reported on a chat I had with a technician of an important automotive company. In that conversation, he sustained that the introduction of sensors, algorithms, simulations and machine learning tools, would improve the way shot peening is applied, both in terms of performance and quality, guaranteeing a constant improvement of the process and growing up the production rate. At the same time, he concluded that he was probably a dreamer since he was not able to see when this would be feasible. Now we are in the year 2024, and the words "Artificial Intelligence" (AI) and related terms are popular and can be found every day in newspapers, magazines and in scientific papers, of course It seems that AI is deeply transforming our society and in a few years, we will not be able to live without these tools. Some are even more pessimistic and think that AI will eliminate or reduce many jobs, leaving people unemployed. I personally am not so pessimistic, and I am convinced that AI tools can be important to improve our lifestyle and to improve the efficiency of industrial production, guiding us toward a more sustainable society. But, what about shot peening and AI?

In fact, shot peening is not so different from other machines or tools. Also in this case, AI can be an important tool for improving the process, making it more efficient, reducing the time for the setup and leading to better performance of the treated components.

That's great. However, AI systems need to be fed with a lot of data if we want the results to be reliable. This means that we must be able to collect a lot of data regarding the technological parameters of the plant we are using, and we should know the effects of different combinations of parameters on the effects induced by shot peening and, finally, on the performance of the treatment in terms of fatigue life and strength. 

In other words, we should be able to define and develop a digital twin of the complete cycle of the shot peened parts. And the more data we process, the better the predictions will be. This means that the model can be improved only if we keep on using new data until we reach a sort of "saturation". In other words, those who say that by using AI there is no longer the need of performing experimental tests do not take into account that these systems are designed to elaborate and use a very large amount of data (the big data). If we focus on the performance of the treated parts rather than just on the process, the deep knowledge of the process can be obtained only by recording what comes from the real shot peening treatments in terms of surface finish, residual stresses, microstructural changes and fatigue behaviour. And we know all how much variables contribute to the final result!

This means that the way we approach shot peening should change to maximize the advantages offered by AI tools, and that shot peening plants should be rethought.

Some innovations are already coming to the market, others will come soon, I am sure.

And, thinking back to my conversation with the technician I mentioned at the beginning of this column, the dream is becoming a reality.

Contributing Editor MFN and 

Full Professor of Technical University of Milan

20156 Milan, Italy

E-mail: mario@mfn.li