How machine learning can optimize the steel rolling process

steel-manufacturing-plant

Rolled steel is a versatile material with incredible physical properties and, as such, is in just about everything: Trains, planes, buildings, and infrastructure. Fan blades, frying pans, and filing cabinets. It’s also a key input in the industrial and manufacturing sectors.

As a $2.5 trillion industry, there are few things that steel doesn’t touch, but COVID-19’s impact on supply and demand has put margins under pressure.

With only a slight recovery forecast for 2021, steel mill operators and machine builders should consider how to protect their margins in a highly competitive industry with constantly changing quality requirements.

Teaching an old dog new tricks

There’s a growing global trend towards the “Smart Mill”, with operators and machine builders turning to data-assisted decision-making to help them react to and prevent production challenges.

Radically more cost-effective than building new steel rolling mills, Smart Mill technology, such as Expert Execution Systems (EES), can optimize the milling process, resulting in reduced operating costs (elimination of the cost of non-quality, for example), healthier bottom lines, and a smaller carbon footprint.

However, to understand where the industry is going, it’s useful to know how it’s evolved.

The best analogy for this digitalization trend and the evolution towards autonomous manufacturing we’ve seen, is that of the autonomous car.

Journey towards autonomous manufacturing

Everyone has a basic understanding of how a car works. With the right combination of clutch control, acceleration, steering, and braking, the vehicle moves from A to B.

Usually, there’s only one person in control of this process, and – most of the time – there are no major incidents. But imagine a scenario where, instead of one person controlling everything, there are a bunch of people each controlling one thing: One passenger controlling the steering wheel, another the brakes, another the clutch, etc.

For the car to move, every passenger has to execute their assigned task at a precise moment. If the person controlling the gears doesn’t communicate with the person controlling the clutch, who doesn’t communicate with the people controlling the foot brake and handbrake, the car might stall, or worse…

The manufacturing environment today resembles this scenario – a bunch of different processes and systems working to achieve a single goal, but not always executing at the precise set-points, which can result in scrap, downtime, and defects.

While each of the manufacturing parameters may be within their individual tolerance limits as a stand-alone single process, it’s the complex interdependencies and interactions between processes that introduce risk in a multi-step manufacturing process.

Enter artificial intelligence and the rapidly advancing journey towards autonomous manufacturing. With AI, manufacturers are able to understand the complex relationships between historical and current process variables. It then offers holistic guidance and prescribes optimum control parameters to improve production and create a single system working towards a common goal – i.e. getting the car and its passengers from A to B, safely, automatically, and with the least amount of wear and tear.

Autonomous cars will use AI and machine learning to understand the complex interactions between these processes, in a sense, centralizing the decision-making as if one person (or machine) were in control.

To do this, car manufacturers need to understand how a change in one process impacts the rest – in real-time. The problem with real-time, however, is that it’s sometimes too late because it ultimately remains a reactive paradigm where systems engineers have sought to reduce the response time to as short as possible.

The same is true for steel rolling mills the world over: automation is already widespread; what’s missing is an Expert Execution System to make the automation contribute towards the reduction in operating costs.

How AI optimizes the steel rolling process

steel-manufacturing-process

Like cars, steel mills need hundreds of processes to execute at precise moments to achieve the desired output, with minimal defects and downtime.

Sensors on the factory floor continuously record data on parameters like roller speed and pressure, metal thickness, line tension and velocity, temperature, and coolant flow. But without an understanding of the complex interactions between these parameters, and how a change or failure in one component impacts the rest, the mill is just one machine set point away from producing scrap.

Even if a problem is detected in real-time, you’d still need machine builders with an advanced understanding of the physical process to troubleshoot and find the right machine set points. It’s a time-consuming process because it only focuses on one sub-step at a time, while costs from factory downtime tick higher.

Enter Expert Execution Systems.

The parameters in steel rolling mills are controlled by systems that are trying to achieve set points. With potentially thousands of set points across machines, finding a configuration that produces the best quality rolled steel output requires the expert knowledge of machine builders.

Unless you have an Expert Execution System (EES), that is.

By analyzing historical events, the AI-guided EES learns – very quickly – how a change in one set point impacts the final quality of the rolled steel.

It becomes an expert with experience in all of your production history and can then prescribe optimum plant control parameters, holistically,  to minimize downtime and eliminate defects and scrap.

With an understanding of “what good looks like”, AI algorithms use feedback loops to monitor, analyze, and optimize critical parameters. EES continuously prescribe changes to set points to ensure overall equipment effectiveness. 

The result? 

  • Massive cost savings in the form of reduced energy consumption per steel roll or beam produced.
  • The ability to optimize process parameters ahead of real-time, i.e. before scrap happens. In our experience, EES reduce the cost of non-quality by an average of 40%.
  • Reduce or prevent surface defects, such as dents, scratches, and scale residue.
  • The ability to strictly manage quality criteria without the expert human analysis that is typically required.
  • Increased production capacity through the reduction of scrap.

In eliminating the “physical mode”, EES reward machine builders with more time and flexibility to expand their product ranges and to offer their expertise and value along the entire steel rolling process – not just at the machine level. They’re also able to provide their customers with a clear path to upgrade their machines and processes – using data they already have.

DataProphet PRESCRIBE is a proven, cost-effective EES that uses AI-as-a-Service and machine learning to optimize the steel rolling process.

Using feedback loops and real-time analysis, DataProphet PRESCRIBE lets operators and builders accurately schedule maintenance planning and production line scheduling, while improving overall equipment effectiveness.

What’s more, DataProphet PRESCRIBE:

  • Connects to existing process and quality databases, including Roll, Pass & Planning Databases, Process Historians, MES systems, and quality databases – even manual inspection logs.
  • Integrates into existing machinery, allowing mill owners to modernize plants and operate old equipment like new.
  • Scales across plants and uses state-of-the-art AI guidance technology to perform automated quality inspections.

With DataProphet PRESCRIBE as part of your AI and digitization strategy, you can unlock ahead-of-real-time guidance on process control changes to prevent scrap before it takes place, while increasing recoveries and improving throughput.

To find out more about how DataProphet PRESCRIBE can optimize the steel milling process, get in touch with us.

Leave a Reply

Your email address will not be published. Required fields are marked *