Digital Transformation in the Metals Industry: From Data to Value

For metals producers, improving digital maturity is becoming increasingly important for competitiveness, quality and production reliability. Successful digital transformation requires a clear understanding of the current production environment, the challenges worth solving and the data available to address them. It also means considering whether the insights gained can actually be translated into changes in production and choosing technologies and analytics that support those changes.

Digital transformation starts with understanding where you stand

The starting point for digital transformation in the metals industry can look very different from one production environment to another. There can be significant differences between production processes, individual plants and even business units within the same company. In some production environments, relatively little data is still collected, while for example highly automated mineral processing operations may already collect extensive amounts of data and use it to develop their own machine learning models.

The amount of data available, however, says little about a plant's digital maturity on its own. What matters more is what data is being collected, how reliable it is, where it is stored and how easily it can be accessed. Ultimately, the question is whether the data can actually be used to improve production. If there is currently no available interface/API to PI historian storing the historical data for training the models, you might as well want to grab a cup of coffee and today’s newspaper because building a functioning pipeline between historian provider and data analytics tool can take up to several months of building, accounting also competitive tendering of service providers and data uploading.

The gaps often become visible only when the data is needed for something new. A process may have extensive continuous measurement data, while production incidents, deviations or operator actions are still recorded manually in some cases. The measurement data may clearly show that something happened, but the explanation for why it happened may be buried in a separate shift log or, in the worst case, a paper notebook. It is easiest is to learn from the mistakes of others, so if you have ‘til this day recorded your unexpected production breaks and their suspected reason in a notebook, now is the perfect time to start implementing a standardized reason code and time-stamping the production breaks alongside whether breaks are unplanned or not.

Fragmented data is another common challenge. Information may be spread across systems that do not communicate with each other or stored in a third-party provider's system. A company may only discover when it wants to make broader use of its data that extracting it is difficult or that access cannot easily be provided to an analytics partner.

Understanding your starting point therefore means looking beyond how much data is being collected. It requires a realistic view of what data is available, where the gaps are and whether the existing data and infrastructure can support the production improvements the business wants to pursue.

Start with a production challenge, not technology

Digitalization is most useful when it begins with a clear understanding of where value is currently being lost, what needs to improve and whether the problem is significant enough to solve. In metals production, the objectives are often same as generically in manufacturing. Producers want higher yield with lower raw material and energy consumption, consistently high product quality and less off-spec production. Production reliability matters as well: the process needs to run as predictably as possible, with fewer disturbances, interruptions, and the costs that come with them.

Broad ambitions such as “improve quality” or “make production more efficient” are not yet a strong basis for a digitalization project. The objective becomes much more useful when it can be expressed in concrete, measurable terms – for example, reducing off-spec production by X percent. That makes it possible to build a credible business case and define KPIs that can later be used to assess whether the project delivered a meaningful result.

In practice, interest in new technology or AI often appears before the company has decided what problem it wants to solve. The starting point can effectively be “let's see some AI”: the technology looks interesting, and only then does the discussion begin about where it could be used. The result may be an interesting technical exercise, but for production teams, it can amount to little more than new buttons and graphs that add work without helping them control the process. Without a clear practical benefit, there is little reason to make the solution part of day-to-day operations.

There should be a clear connection between the business objective, production KPIs and the analytics used to support them. A commercial or production objective can be linked to a KPI and from there to measurable process factors that can influence that KPI. Understanding those relationships also makes it easier to determine what data, analytics or technology can genuinely contribute to the desired outcome.

Defining the objective is a team effort. Management or project leadership needs to understand the expected business value, production experts need to understand the physical reality of the process, and IT needs to know what the existing data and technology infrastructure can support. Bringing these perspectives together helps ensure that the objective is both commercially relevant and realistic in the production environment.

The production problem may start somewhere other than where you see it

When a production target is missed, the explanation is rarely found by looking at a single measurement. Metals production is influenced by several factors at the same time, including raw material properties, equipment condition, process conditions, production rate, and the way the process is operated, among other factors. Raw material variability is particularly important, as its effects can carry through subsequent process stages and ultimately affect final product quality.

The different stages of a metals production process are closely connected. The output of one stage often becomes the raw material for the next, which means variability introduced in one part of the process can travel further down the production chain. Take for example nickel sulfate production: a slightly too drastic pH deviation can change how selectively nickel and impurities are separated. Deviation moves through solvent extraction (SX) stages, amplifying and causing insufficient separation across extraction cells and finally affecting product quality or causing recovery problems.

“In my experience, the problem very often starts somewhere other than where it eventually becomes visible.”

Leo Bruk,Senior Process Expert, SimAnalytics

Tracing a production problem back to its source often requires examining data from several process stages and systems together. Process historians, laboratory systems, production management systems and maintenance records may each hold part of the explanation, but the connections between them are not always easy to identify. Even when the relevant data is available, the time delay between a process deviation and its consequences can make root cause analysis difficult. Again, if we think about nickel and cobalt extraction, impurity concentrations causing decreased Co/Ni separation may only be briefly visible in the DCS, but the detailed process data is stored in the historian, while the effects of the impurities are later reflected in laboratory analysis results. Machine learning software resembles at this point a bit irritating property manager digging around old account statements from the housing company’s history and forcing the board to make more responsible decisions in the future.

Understanding production still relies heavily on experience. Many plants have individual experts who know the behavior of the process so well that they become the people everyone turns to when something difficult happens. Communication between production areas and shifts matters just as much, because a change observed early in the process may become critical information for someone working much further downstream.

This is where broader use of production data becomes valuable. Combining process knowledge, historical data and the experience of production experts makes it possible to investigate the factors that genuinely explain process variability, unexpected production losses, or final product quality problems. It can also make cause-and-effect relationships visible as they are happening, when they would otherwise remain hidden between separate systems, production areas or people.

The right technology depends on the problem you need to solve

There is no single analytics solution that fits every production challenge. The right approach depends on whether the issue is missing or slow measurement data, poor process predictability, or relationships between process parameters that are difficult to understand. If an important variable cannot be measured directly, or the measurement arrives too late to support production decisions, for example, a soft sensor can estimate the variable using other available process data.

Traditional process analysis can solve many problems, and more advanced technology is not automatically the better option. However its limitations become more apparent when production is influenced by a large number of interdependent variables and the relevant cause-and-effect relationships are difficult to identify from historical data. Advanced analytics can help examine these relationships more broadly and monitor how the process is behaving while production is running.

The desired outcome also affects what kind of analytics is useful. When improving quality, the focus may be on understanding the combined effects of process parameters and predicting final product quality, while improving production reliability may involve detecting deviations in equipment or measurements. In cyclical production, comparing the current run with previous production cycles can help identify deviations while production is still in progress rather than discovering them later in historical data.

Even the right technology has limited value if its results remain separate from day-to-day production. Analytics needs to support the people operating the process at the point where decisions are made, whether that means identifying a deviation, adjusting a process parameter or investigating a developing problem. Making this work may require changes to existing workflows so that new information becomes part of how production is monitored and managed. The value of analytics ultimately depends on whether the insights it provides lead to actions that improve production performance.

Digital maturity lays the foundation for future competitiveness

Pressure to make better use of data in the metals industry is coming from several directions. Raw materials and energy need to be used more efficiently, production needs to remain cost-competitive, and requirements around quality, traceability and sustainability reporting continue to evolve. Meeting these demands becomes increasingly difficult without sufficient, reliable data.

A plant's starting point has a major influence on how quickly it can develop its digital capabilities. Where data is already collected extensively and the infrastructure supports its use, new applications can be built on an existing foundation. In an older production environment, simply enabling the necessary data collection may require substantial investment in equipment and infrastructure, raising the threshold for further development.

“What really matters is what you are able to do with the data. Once you have a lot of it, the question is how you use it and turn it into an advantage over your competitors.”

Saara Abdalla, Process Expert, SimAnalytics

For metals producers, the next step depends on where they stand today. Some may need to improve data collection and accessibility, while others are ready to use advanced analytics to address specific production challenges. In either case, progress comes from connecting the right data and technology with production expertise and making the resulting insights part of everyday decisions. That is how digital transformation can deliver measurable improvements in production and strengthen long-term competitiveness.