FOLLOW US Twitter CONTACT US FTJ Email address Phone number

Welcome to the digital foundry – powered by AI

Spanish foundry group Condals cut scrap by 45 per cent after implementing DISA’s Monitizer® digital solutions, with customised dashboards and Artificial Intelligence-driven recommendations transforming the company’s digital capabilities, writes Kasper Paw Madsen, DISA, Taastrup, Denmark. At its two Spanish and Slovakian locations, Condals Group’s three DISAMATIC® moulding lines produce over 43,000 tons of iron castings each year. To improve casting quality, efficiency and productivity, the company employs traditional good foundry practice like modifying patterns or varying mould density. But these days, much of its insight arrives through collecting, visualising, and analysing process data. So, when long-term supplier DISA added an Artificial Intelligence (AI) product to its Monitizer suite of Industry 4.0 solutions, Condals was naturally curious.

“Our goal is to be a data-driven company,” says David de la Cruz, CIO at Condals Group. “In 2020, we were interested in finding an AI solution that would help us lower our scrap rate as much as possible. Which is why we spoke to DISA about Monitizer | CIM, GLOBAL and PRESCRIBE.” The Monitizer Suite supports any foundry at any stage of its digital journey, whether they are novices or experts. Using Monitizer | CIM with DISA machines, individual foundries can collect data, implement automatic process control and perfectly synchronise their DISA equipment – cutting out the need for manual intervention.

Monitizer | GLOBAL collects data from any vendor’s equipment across multiple foundries, lines or sites, and presents it in real time to users who can monitor and analyse it, then use the learnings to inform improvements. Monitizer | PRESCRIBE, powered by AI partner DataProphet, takes an even bigger leap forward, with automated AI-driven analytics that deliver dynamic, real-time process analysis across an entire production line to significantly reduce scrap and improve profitability. “DISA offered a complete solution with Monitizer,” explains de la Cruz. “We already had Monitizer | CIM installed and that had worked well for us. Monitizer | GLOBAL is a really powerful tool for putting all our data in one central place and making it easily available.” In November 2019, Condals decided to implement the entire Monitizer Suite. It was time to create a next-generation digital foundry.

Building on a long-term relationship

Condals has two DISAMATIC® lines at its main Spanish plant – one installed over a decade ago and another with a state-of-the-art DISAMATIC D3 vertical moulding machine. There’s a DISAMIX S100 sand mixer and another D3 in the single line in the company’s Slovakia foundry which opened in 2016. Condals has invested in DISA’s accessories too, such as the Mould Accuracy Controller (MAC) which arrived in 2020. “We have been working with DISA since the beginning of our company and their equipment has helped us to be successful,” says de la Cruz. “We already have a lot of data and we wanted to get more value out of it and for our people to have one single source of information.”

The project team first upgraded Condals’ existing Monitizer | CIM installation. CIM already supported tasks like recipe management, in-mould cooling time control, automatic pouring and automatic casting sorting. The latest version would make reliable time-stamped, structured data available from all the DISA equipment, improve machine automation and recipe management, and support modern communication standards plus a wider range of languages. Monitizer | GLOBAL came next, bringing all Condals’ data together in a unified cloud-based database. The company previously had to dig into each system separately, then merge data sets and calculate derived variables manually; GLOBAL’s automated variable calculations and data integration would make most of this processing unnecessary. “Because we have multiple different systems with different databases, our data was split into silos,” explains de la Cruz. “We mostly had to aggregate and integrate it manually. To gain a view of the whole process, we would have to spend days combining and analysing data sets.”

Information is power

Condals wanted access to real-time foundry data to be able to respond immediately to any line issues and also wanted that data stored in its data warehouse. “Monitizer | GLOBAL will give us real-time data to work with right now that is also stored, so we can work with it afterwards,” says de la Cruz. “We also wanted a single standard source that everyone could trust and access. Before, the same analysis could give different answers depending on how the data set was brought together and the time period under consideration.” The DISA team audited the available data and identified the sources, then used NoriGates (Edge Gateways) to securely collect and transport data to GLOBAL’s cloud database from a wide range of sources across Condals’ three lines. All deployment, configuration and testing took place during the Covid pandemic, so Monitizer’s remote access capability was vital. “We were only able to meet face to face twice,” explains de la Cruz. “We consulted remotely with DISA’s engineers and we fitted the NoriGates in Slovakia ourselves, which was very easy. Considering the global situation, the whole installation went amazingly smoothly.”

Making sense of complexity

Now data access is easy and instant. Instead of struggling in Excel to compute important metrics like cumulative totals of iron poured, Monitizer | GLOBAL automatically calculates and displays these derived variables in real time. As of November 2020, Condals had twelve systems connected to Monitizer | GLOBAL in Spain, with almost 2,000 different data points available through over 40 different dashboards. In Slovakia, there are currently two systems connected supplying over 500 data points to six dashboards. Dashboards in both locations display real-time metrics, customised for each user. For example, a machine operator might see the current line speed while managers might prefer average speeds over the last hour, the last shift or the previous day. One popular dashboard option displays all the important process data related to castings created from a single pattern, with an ‘end of batch’ report summarising each run. “Now anyone can connect and see the data from anywhere,” enthuses de la Cruz. “We can collect and combine the data much faster than before and that makes analysis faster too. Managers, supervisors and operators, many now have their own personalised dashboard view based on their role. We configure the dashboards and KPIs on our own most of the time, with DISA always ready to support us remotely.”

Towards zero scrap

With the data foundation in place, it was time for Monitizer | PRESCRIBE and the AI phase of the project. Testing started with a single Spanish line. “Monitizer | PRESCRIBE relates all the process parameters from the entire production line to one output – casting quality – and we found it was a very easy solution to understand and use,” says de la Cruz. “Other competing solutions only look at parts of the process in detail and only optimise certain parameters – mould alignment, temperature, porosity – whereas Monitizer | PRESCRIBE optimises the entire process to reduce scrap. That was exactly what we were looking for.” The PRESCRIBE project team extracted, transformed and loaded Condals’ historical production data into the Monitizer | PRESCRIBE platform. Over 700 parameters described the processes for the initial five test patterns. Multiple methods are used to track molten metal and individual moulds and castings, such as mould numbers and pattern keys. Where no tracking method exists, software tracing techniques can detect process events like molten metal leaving the holding furnace or mould pouring, and calculate the time taken to move between them. For example, a spike in pouring ladle temperature and weight would mark exactly when the ladle was filled with metal. Once the unified, time-synchronised data set was ready, model creation and training began. Through advanced, unsupervised machine learning, the AI’s neural network calculated and established the correlation and interaction between the hundreds of input process and machine variables, and final casting quality data. This produced two initial models, each of which would automatically specify the optimal operating regime for one of two test patterns. After rigorously checking the models’ predictive accuracy, real-life commissioning began.

Inspiring improvements in casting quality

As the prescriptions went into operation, improvements arrived quickly. One test pattern already had very low scrap rates, but that was reduced by a further 39 per cent. The other pattern had a high existing scrap rate, which the prescriptions helped cut by 45 per cent. “These initial commissioning results are very encouraging,” states de la Cruz. “Though we are still at an early stage, I think this is the right approach and we are going in the right direction. We definitely expect further improvements in future.”

Not only did Condals see impressive results rapidly, but the improvement has since been maintained. “The scrap rate is staying at the lower level and actually still dropping slowly, so we have seen a reduction and then a stabilisation,” explains de la Cruz. “The prescriptions are pointing us in the right direction, for example, in showing us which variables have the most influence on our process and so what our priorities should be.” The focus is now on fine-tuning model performance and implementing the advice in the prescriptions to move the process parameters into the desired optimum operating range. Prescriptions are currently delivered, reviewed and implemented weekly by a dedicated in-house team, but the aim is to move to real-time working as soon as possible. When that happens, machine operators will see recommended machine settings pop up in their dashboards. “We don’t want to change everything we have been doing for the last 20 years immediately, but will gradually make the prescriptions part of our process,” explains de la Cruz.

Accelerating into a digital future

Condals has already seen major improvements in its digital capabilities. “We have certainly been able to find things that make us more profitable and our process more understandable,” says de la Cruz. “The great thing about PRESCRIBE is that it brings everything together to predict its influence on scrap and gives you a clearer picture of what is really happening. The journey is really important – what we learn along the way to gain greater understanding of what affects scrap.” The models that have been successfully tested are now in live production and the next phase of testing has already begun. Data is now being collected for 38 patterns, with extra data sources added to make the predictive modelling and prescriptions more accurate and present an even richer real-time view of the entire line. Once PRESCRIBE is fully proven on the first test line, Condals will extend it to its other Spanish line. “I can for sure recommend the complete Monitizer package,” concludes de la Cruz. “We have been able to unify all our data in the dashboards and that’s a clear advantage. Monitizer | PRESCRIBE is definitely the right approach, it shows us the key ways to reduce the scrap rate and is already giving us excellent feedback.”

Kasper Paw Madsen is the global product manager for digital solutions at DISA.