An interview with Alberto Albertini - Innovation Center Director, Technology Scouting at Antares Vision Group
Metallurgy: aluminium die-casting from data to action (and savings).
1 December 2021
Reading time: 6 minutes
Hi Alberto! Our blog is popular mainly among data scientists, so a brief introduction is necessary: what is high pressure aluminium die-casting?
I start from the basics due to my thirty years experience in this field. Die-casting is the process of injecting a molten metal (mainly an aluminium alloy) at high pressure (hundreds of bars) into a steel die. Once the metal is solid, the machine operator opens the die and extracts the casting.
It sounds simple, but a machine (we are talking about horizontal and hydraulic machines, usually called “die casting machines”), has an extraordinary number of parameters to regulate and monitor in order to work properly, and these values are not stable over time, but must be continuously modified during the process, also due to the interaction with other automation peripheral devices.
In addition, we are talking about the production of aluminium and magnesium alloy castings with extremely challenging designs and very limited thicknesses, especially for automotive or electronic components. Die design and engineering is crucial for the success of the casting, as are a lot of other inter-related parameters.
In order to provide an idea, according to a study for a machine manufacturer from close to one thousand parameters available in the machine interface, around 28 parameters influence scrap, 15 parameters impact machine availability and 6 parameters affect the overall performance. You can imagine how challenging it is for an operator to optimise machine performance without the help of data analysis and optimisation tools.
Talking about data-driven innovation, what is the state of the aluminium die-casting industry?
Over the last 15 years, this sector invested a lot of resources in automation, especially in the “Automated Work Cells” through the integration of additional machines and related electronic monitoring and management systems. Over time, the mechanics (except perhaps for the machine with two platens instead of three), the hydraulics and, above all, the on-board electronics evolved very gradually, while the increasingly pressing goal became to make the process more efficient and high-performing. A considerable effort was also spent on process optimisation and cycle time reduction, increasing the dies’ productivity, in order to produce larger castings with minimum thicknesses (the “structural castings” forming the car chassis and body).
On the other side, Computer Vision and Data Analysis remained in the background.
In your opinion, what is the reason for this lack of data-driven innovation?
Computer Vision and Data Analysis are new topics for all industries, and definitely the die-casting sector has been more conservative and traditional for decades, as well as the entire manufacturing industry is still far from other sectors where data are more easily collected and automatically analysed.
In this sector there is still a strong dependency on the operator and his/her knowledge of the specific machine and process. But it is also a sector which lacks young and trained operators able to facilitate an effective generational turnover.
More skills is what we need to deal with in order to bridge the gap. Machines are constantly evolving in terms of technology, but what about people?
The educational activity carried out by the CSMT of Brescia (https://www.csmt.it/) moves exactly in this direction: the High Pressure Die Casting School and the Low Pressure Die Casting School, started in 2014 and 2017 in collaboration with AQM, arise to answer the market need, in particular for the Automotive, by technically training the future professionals to ensure the operability and competitiveness of our companies.
On the other side the educational programs do not cover all the newest technologies, also because artificial intelligence is quite recent and still does not contain an adequate offer of consultancy, programmes and services for companies. In this sense, Oròbix established a working method that places companies at the heart of the AI adoption process, fostering the transfer of expertise in order to increase the awareness of new technologies.
Regarding data, in your opinion, which applications are most ready to be deployed? As you know, at Oròbix we're all about facts...
There are many applications, I would split them according to the input data analysed: firstly, the data streams directly from the process, and secondly, the images, for example, the thermographic images of the dies. Die-casting is a process where temperatures play a key role and they are not still properly considered (together with the aluminium alloy properties, the “raw material” check).
With regard to the data from the process, they could be used primarily to set up the machines. In fact, there is no standard and constant calibration, the parameters are not stable over time but change in relation to multiple variables, leading to castings not always meeting the quality standards. Anomaly detection tools can be useful for identifying in advance any deviations in parameters, avoiding the production of non-conforming parts and the waste of resources related to the following activities (e.g. machining).
An intelligent real time machine monitoring system could be a powerful tool to improve the operation of machines, with significant improvements in OEE (Overall Equipment Effectiveness). In order to give an idea of the achievable results with this type of system, we can estimate an improvement in OEE – which is the main indicator for measuring machine efficiency – greater than 10%. Considering that the margin of a die-casting machine is often much less than 10%, these numbers should generate significant interest.
The same tools can be used for predictive maintenance on machine components, especially electro-valves (which are increasingly powerful in terms of feedback, to ensure real-time commands and retro-actions), allowing the prevention of failures and unexpected machine downtimes and the planning of maintenance operations, based on real time data and not on the generic technical data sheet indications, with consequent and considerable benefits also for the management of the spare parts warehouse and assistance service.
The same data can also be used to “train” an optimisation model for the machine parameters, which learns – on the basis of previous experience – fine-tuning strategies to get proper castings, reducing cycle times and maximising productivity.
Finally, it is possible to work with computer vision algorithms, using thermographic images that show the hot and cold regions of a specific machine and die, in order to detect in advance any drift or abnormal phenomena and to give instructions to operators and thermo-regulation controllers to return the system to its normal state.
So, once again, data will save the world of aluminium die-casting...?!?! 😁
As a ‘die-caster’ (‘fondeur’, they say in my area, probably evoking some kind of French influence, considering that in my valleys people working with metals for centuries), I still remain realistic: the ambition is to understand the die-casting process a little better with data, but we can’t presume to do this without the help of experienced specialists.
However, there is no doubt that data analysis can highlight critical aspects of the process and give operators the opportunity to react in time. The results could be surprising for the sector.
I believe there is an enormous opportunity to help operators in the field to learn how to manage the machines not only through experience but also through the outputs of the algorithms, saving a lot of time and allowing them to spend time and resources on more specific tasks and overall monitoring. An extremely useful magnifying lens on what is going on.
What would you suggest to an entrepreneur in this sector looking to understand more and maybe start an artificial intelligence project?
First of all, don’t think that artificial intelligence is something academic, limited to specific sectors, or even that it only works in the labs. There are several practical applications in manufacturing, across the entire supply chain, which give credibility and consistency to these techniques.
And then my suggestion is to start from an ‘entry level’, for example a very specific problem, even a small requirement (but which could have an important impact on the cycle time or on the quality of the finished casting), approaching it with tools that can bring the potential of artificial intelligence applied to the industry and to the data: starting with small problems and then scaling up to more complex ones.
In addition, it is important to allocate human resources to these activities because results will only be achieved through mutual and continuous exchange between data scientists and process operators. Understanding and interpreting the results is crucial. Providing technical staff with artificial intelligence tools is an investment in competitiveness that will quickly give a return.
And to our data scientists working on manufacturing projects, what do you suggest?
To listen to the process operators and to trust their experience and expertise but at the same time not to be scared of breaking down the assumption I have heard so often: ‘we have always done it this way’. I believe the additional value is the attention Oròbix has for the operators, ensuring their involvement in the whole process. Working together, listening to the operators and relying on their experience and expertise. Sometimes it will be difficult to overcome natural resistance, and some prejudices, especially in a consolidated and traditional industry such as die-casting. Data can open up new ways but in order to be considered it’s necessary to make the analysis as informative and clear as possible. The value for business, or more importantly the tangible savings, must be clear and never look like statistical exercises for their own sake.
Alberto Albertini was born in 1966 in Brescia, where he lives. Graduated with honours in Modern Philology, he has been working in the manufacturing industry for 35 years, 29 of which in diecasting, and today as Head of Innovation and Technological Scouting for Antares Vision Group.
Copywriter, journalist, marketing and communication consultant, lecturer at the Faculty of Linguistic Sciences of the Università Cattolica del Sacro Cuore in Brescia, one of the founders of the art magazine Stile Arte, he collaborates with the Scuola Holden and newspaper Il Giornale di Brescia, and he is the creator and artistic director of the Rinascimento Culturale festival. His novel “La classe avversa” (Hacca Edizioni) was one of the most awarded Italian literary debut of 2020.
Download the Detectiv.ai brochure and the case study.
To know more about our work, check out our AI solutions page. For any further queries or demo request, feel free to contact us!
Interview with Lisa Lozza -Technical Area Leader Data Science