Published

The next step in digital transformation: is Artificial Intelligence production-ready for green sand foundries?

A new digital solution from DISA and DataProphet offers foundries a practical and proven way of using Artificial Intelligence (AI) to optimise their processes and reduce scrap. We talk to Kasper Paw Madsen, Global Product Manager for Digital Solutions at DISA, and Frans Cronje, CEO of DataProphet, about how AI is already helping foundries move quickly towards a zero-defect future.

Kasper and Frans, thank you for joining us today. To kick off, can you tell us – briefly – why using Artificial Intelligence (AI) in a green sand foundry is a good idea?

Kasper: DISA has been helping foundries collect, visualise and analyse their data with our Monitizer® suite for a few years now. Adding AI capabilities to do more with this data is a logical next step – and it’s a big one. Monitizer | PRESCRIBE – which is what our AI product is called – harnesses the power of AI to optimise the whole foundry process, significantly reducing scrap while increasing capacity and production predictability.

Frans: There’s a lot of hype around AI so at DataProphet, we like to quote real results to show what’s possible. Over the last two years, the average AI-driven defect reduction across all of our manufacturing customers is 40%. With some, it’s 80% or 100%. Few foundries take full advantage of Industry 4.0 techniques so the potential for them is enormous.

Our Expert Execution System (EES), enabled by AI, has helped a foundry in South Africa cut defect rates in grey iron engine block castings by 50% in the first month. For the first time ever, they achieved zero internal defects on all shipped castings over three months and save over $100k every month.

How does AI help deliver these kinds of results?

Kasper: The key word here is automation. Many green sand foundries already collect and analyse process data but it’s usually limited to single sub-processes like moulding or pouring. The data for each process stays separate and basic manual analysis is done using spreadsheets or simple statistics.

With an entire foundry line, optimisation can involve hundreds or even thousands of variables across all the different process stages. Making sense of that complexity manually is just impossible. AI automates this analysis, using the cloud to access vast computing capacity. That’s the only way to handle the complex, large sets of data that will give us new insight that will in turn make a genuine difference to a foundry’s performance.

So what does an AI solution like Monitizer | PRESCRIBE actually do?

Frans: It starts by analysing historic production and quality data to learn from past mistakes and corrections, to find what works and what doesn’t. It considers how the parameters within and across all the different processes are related, how each one influences the other and what the ultimate combined effect on quality is.

From that analysis, Monitizer | PRESCRIBE finds the optimal process parameters and tolerances for a particular casting and process. Knowing the best “recipe”, it can prescribe – hence the name – the best actions to take to improve quality.

Kasper: A good example is where, even though all your process parameters are within tolerance, you still might see bad quality castings. Often, this is because one metric is slightly high, another is slightly low and so on. It’s a specific combination of values that produces the defect, not a single extreme one.

Because the AI has learnt how parameters like grain size, moisture content, pouring speed or inoculation rate influence each other, it can pick the right settings for minimum defects.

So that’s like a much more effective version of today’s offline analysis. How does AI help you apply those learnings during real production?

Kasper: Monitizer | PRESCRIBE applies what it has learnt to live data – keeping an eye on what your foundry is doing right now, in real time. That gives you dynamic process control, reacting instantly as conditions change, like ambient air temperature or sand moisture content, and telling operators on the line the optimal settings or actions to take in time to prevent defects occurring.

It keeps on learning too, constantly optimising the production process towards zero scrap and improving other metrics like productivity and resource use.

Frans: Data-driven, real-time optimisation is sophisticated second-order control. By constantly monitoring machine and process data, then telling you which adjustments to make and again monitoring their effect, our AI tool gradually gets every part of your process running in harmony. You achieve a stable operating regime with the best quality and minimum quality variance.

A good analogy is with an autonomous car which can automatically keep you in the middle of a motorway lane. By constantly computing the optimum process parameters, our AI keeps your process in the “middle of the lane”.

It’s clear that automation and data analytics have enormous potential but many foundries have yet to adopt the basics here. So is it really possible for any green sand foundry to make use of AI?

Kasper: We see digital as a four-step journey where you start with data collection and visualisation, then move at your own speed towards analytics, AI and automatic process control. Of course, we can help customers do all of that very quickly if they want to.

Our NoriGate is the only hardware involved – for data collection – and everything else is a cloud service which we can deploy in any foundry or with existing data collection infrastructure. That makes it very quick and resource-efficient to deploy. You won’t need any new IT hardware or to employ data scientists or any extra staff.

We can digitise every step in the green sand process, take data from paper records or pull it from Excel, and give you a single trustworthy, time-stamped database ready for investigation. At each step, you can achieve significant benefits.

The point is that, no matter if you are just starting out or are digitally advanced, there are things we can do that help you take the next step very rapidly indeed.

So you don’t have to be a rocket scientist to make use of AI?

Frans: AI’s inner workings can be complicated to understand but together we have developed it into a packaged service that works for foundries. It’s not hard to implement it and it’s not capital-intensive. As Kasper says, everything you need to collect, store and report on the data is already available from DISA and well proven.

Some foundries think they are too “old school” for digital, but AI projects can be realised when there’s no strong data environment or even if they haven’t really previously captured data at all. Our partnership with DISA enables very rapid digital progress in any type of foundry.

Does your partnership – between an industrial AI company and a foundry equipment expert –make your solution different to the other AI products we see emerging?

Kasper: A lot of vendors say they have an AI system, but a pure IT company may never have seen a foundry from the inside before. We bring a combination of deep foundry experience and DataProphet’s award-winning expertise in manufacturing data science with more than 35 engineers, statisticians and computer scientists dedicated to developing AI solutions.

 

This collaboration makes our service uniquely practical and effective. It’s already tried and tested in a green sand foundry environment and we’re finding that fact is very attractive for customers. For example, we are currently installing the full Monitizer suite including Monitizer│PRESCRIBE at a large European foundry group.

From DataProphet’s point of view, how does DISA’s experience in green sand foundries help an AI project succeed?

Frans: When you implement an AI solution in manufacturing, it’s vital to capture domain knowledge completely and correctly. As the leading OEM supplier, DISA know green sand intimately and are very much the experts in the foundry environment. They know what to do and which questions to ask right at the start. That means value from a running system arrives in weeks, not months or years.

DISA’s customers also trust them to keep their promises and they understand that Monitizer│PRESCRIBE will be delivered and managed through them. If DISA puts its name to it, customers know it will be an effective, high quality product and that will be supported in five years time – and in ten or twenty years too.

Is this AI solution just for DISA customers?

Kasper: The entire Monitizer suite, including NoriGate and Monitizer│PRESCRIBE, is machine-agnostic, so it’s not limited to DISA machines or even to the green sand process. Monitizer is a Norican-wide solution, so every foundry can benefit from it, whether it’s pouring iron or die-casting aluminium.

Frans, with your experience, how do you think foundries compare to other manufacturers in their application of digital tools?

Frans: Some other manufacturing environments are now quite sophisticated in their use of software and data, which is not often the case for foundries. With IoT infrastructure and Expert Execution Systems like Monitizer│PRESCRIBE, there is a real opportunity for foundries to leapfrog the older IoT systems and access the very latest technology without having to make an enormous investment.

Are there any common misconceptions about AI you hear from your foundry customers?

Frans: They can be worried that their data might be used in another customer’s AI – which never happens. Monitizer│PRESCRIBE can ingest and interpret all a customer’s foundry data and that certainly doesn’t include data from other customers.

Monitizer | PRESCRIBE is designed with full tenant sandboxing: every client’s datastore, database, and model is uniquely encrypted, and every component is isolated from every other component in the system. It is not possible to mix data or models between clients and the data is safeguarded with every possible measure.

Kasper: Some people think AI needs another in-house IT system that’s big, complex and very expensive. But Monitizer | PRESCRIBE is an online service, it simply gives you a tool to optimise quality and productivity.

Also, when we talk to foundry staff, some fear an AI system will come in and take over their job. But this isn’t about taking jobs. The information AI gives will help them make better decisions and improve their own performance. It will make them look good.

Are there any other AI-related advantages for foundry owners and their workforces?

Kasper: There’s a generational change going on in our industry. Engineers with 30 or 40 years’ experience are retiring and our customers are worried that their knowledge of how to keep their own unique processes running correctly will be lost. But their knowledge is encoded within historical process data. Monitizer | PRESCRIBE can access that and put it to work.

With more automation, the foundry also becomes a cleaner, more attractive place to work. You can spend most of the time in an office-like control room, which will be more appealing to today’s potential recruits.

Frans: By learning from human intelligence, expressed in millions of decisions made over the years, the AI becomes the central knowledgebase for the foundry. Then it can support less experienced engineers and operators in their decision making. A lot of value for manufacturing customers lies in selecting and extracting those good decisions so they’re never lost.

If AI helps foundries move from offline analysis to continuous guidance, what comes next?

Frans: The end goal is a foundry that runs its own processes automatically – similar to what the autonomous vehicle industry is aiming to achieve with cars. Staff will gradually move from continuously analysing processes and adjusting machines to focus on tasks they’re better suited for like innovation, creation and ideation. The plant of the future will re-configure itself for the optimal delivery of new customer orders, adjusting its configuration, production schedule, energy consumption and staff roles to give maximum efficiency.

Kasper: The system will adjust settings automatically, for example, when sand properties change, and you need more additives, or if the humidity changes and the sand dries out faster so you need to add more moisture. All these variations are corrected manually today and, even with Monitizer PRESCRIBE’s real-time advice, usually still will be, but the system will handle it all automatically in future.

How close is this fully autonomous future?

Frans: We’re not there yet, but it will definitely happen for some foundries in the next few years. Most foundries are starting to collect data and analyse it, so they are being assisted by data today. Our system goes from there to guiding them with specific real-time recommendations. The “self-driving” foundry is the next stop on the journey.

Kasper: We’re already helping customers fully automate parts of their DISA line, like moulding and pouring, or sand mixing and moulding, though complete automation of the whole line is a little way ahead at the moment. But I think it will arrive a lot sooner than completely autonomous cars.

Many thanks to both Kasper and Frans for a fascinating explanation of how they are working together to bring AI to foundries.

DISA’s AI solution Monitizer | PRESCRIBE is currently live with selected pilot customers and will be available in the coming months.