A Practical Implementation of Industry 4.0 in Foundries

Jim Wenson, Eric Nelson, Lizeth Medina Balliet

Klaus Schwab summed up Industry 4.0 at the World Economic Forum by saying, “It is the fusion of these technologies [AI, big data, IoT, bioinformatics] and their interaction across the physical, digital and biological domains that make the Fourth Industrial Revolution fundamentally different from previous revolutions––diffusing faster and more broadly than any of the previous revolutions.”
Consider these elements/phases, which are at the core of Industry 4.0:

  • Historical data collection.
  • Live data capturing via sensors.
  • Data aggregation.
  • Feature engineering.
  • Connectivity via communication protocols, routing and gateway devices.
  • Integration with PLCs.
  • Dashboards for monitoring and analysis.

These are the core technologies behind the buzz words everyone is printing on their marketing materials. Having a working understanding of the elements of Industry 4.0 is useful, but it does not help build the roadmap for an implementation strategy. Only leveraging the experience of early adopters and industry partners, as well as thoughtful planning lead to successful adoption of any Industry 4.0 initiative.

Big Data

To begin laying the building blocks for Industry 4.0, data needs to be gathered. Many foundries have already been collecting data for years, but is it the right data? Is it useful? Is it even being used? 
Often, foundries will start data collection on an ad hoc basis. The plan usually starts with an initiative to connect anything and everything they have and see what insights can be found. Each purchase is made with an emphasis on ensuring that equipment is “connected” or can be fitted with a sensor. From SCADA, ERP and OEM-provided dashboards, data is gathered all over the place and is typically difficult to synchronize in a workable fashion. Data is often gathered but never used.

Unfortunately, this unstructured method leads to only going into the data after something has broken and the maintenance department needs to determine what has happened. What they often find is that data collection has stopped, has been overwritten due to space constraints, or simply was recording the wrong information, rendering the data collection to be “Big Data” in name only. This is what is referred to as “Dark Data.”

Data should provide information that can be turned into knowledge, which can help build the intelligence behind making the right decisions. With a solid data foundation, operations can begin to know which levers to pull and which buttons to push to answer some of the most basic but important questions:How do we improve overall equipment effectiveness (OEE)?

  • How do we even track OEE?
  • Where are we losing throughput?
  • How can we minimize downtime?
  • What operation is lagging in our machines?

Data alone is not Industry 4.0. It is the combination of data + software + analytics that equals the platform for Industry 4.0 to provide business value and ultimately actionable information.

The goal boils down to obtaining complete visibility of the processes and ultimately using this new insight to drive costs down with actionable data, predictive maintenance and process optimization. 
Implementation at Two Foundries

Two leading foundries in the industry took similar strategies in implementing Industry 4.0 tools in their businesses. Instead of collecting data on every single piece of machinery or choosing equipment that was not production critical, they chose to move forward in areas with the highest visibility and had the greatest potential for immediate wins. At AFS Corporate Member Dotson Iron Castings (Mankato, Minnesota) this meant rolling out a system that highlighted and displayed its three Sinto flaskless molding lines, from automatic mold machines through the cooling turn tables. Neenah Foundry (Neenah, Wisconsin) brought its two existing Kunkel Wagner and BMD tight flask lines online to receive a clear visual picture of how these machines operate and find opportunities for optimization.

By choosing high profile areas within their plants, both companies were able to quickly see results and apply the data they collected directly to their current processes. Since these lines are essential to day-to-day operations, the employees responsible for maintaining and operating the equipment had high levels of knowledge about machine function and could therefore easily interpret data being collected to validate its veracity. They also had a vested interest in using the data to increase their output and machines’ effectiveness. The ability to quickly correlate data and outputs to real-world operation sped up the adoption process, culture change and internal acceptance of the new tools. There would be no sense of urgency to implement, analyze and make use of big data if the equipment chosen had not been critical. The winning strategies for both a medium-sized “jobbing” foundry and a large semicaptive foundry followed a similar pattern:

  • A champion or project team was identified and charged with ensuring a successful adoption.
  • Critical, high-visibility machines.
  • Areas with greatest potential for wins.
  • IT/OT roles defined.
  • System selection (build vs. buy).
  • Clearly identify success criteria.


Industry 4.0 is a natural convergence of Information Technology (IT) and Operational Technology (OT) (Fig. 2). These two separate areas need to agree on responsibilities in a connected plant due to the largely blurred lines. IT is responsible for data and the flow of digital information. It encompasses everything from IT infrastructure, plant networks and application management. IT is responsible for laying out and implementing standard architectures, data storage and management protocols and procedures, ERP integration with the shop floor, as well as any remote access granted through VPN management.

OT is the operation of physical processes and the machinery used to carry them out, as well as the monitoring of these workflows. OT has now become more closely related to IT with traditionally “offline” machines being connected to a wider network.  No longer is the data generated by these machines quarantined to their HMIs; instead, they can be accessed virtually anywhere. 

With OT becoming more of a connected world, a new set of roles are emerging. Responsible parties must be aware of cybersecurity vulnerabilities through training and take the necessary steps to control these risks. OT must also be responsible for how the machines connect or how they will need to be modified in order to gain that ability. 

Build vs. Buy

Many companies decide to build and maintain an in-house solution customized to their operation. The biggest downsides to in-house development are the hidden upfront costs, implementation and long-term costs. In 2011, PricewaterhouseCoopers research concluded companies routinely underestimate the real cost of building and operating in-house systems. It found that the main reason for underestimating the real value of custom solutions was the failure to account for hidden costs, which typically account for more than 50% of the actual cost. In a purchased system, the development costs, on-going support and feature-enhancement costs are spread across many users rather than a single organization. 

No one has more intimate knowledge of a foundry or plant than the team members who operate it every day. However, having dedicated personnel to develop software is typically not within budget, and those charged with the task more than likely have other responsibilities more essential to the profitability of the organization. Freeing these people to do their “day job” can result in a much more focused and effective team. 

Success Criteria

At Dotson Iron Castings and Neenah Foundry, success was measured through various outputs. They broke down criteria based on who was using the data and posing the following questions:
From an Executive’s Perspective:
How fast can it be implemented?
How do we limit the number of additional devices and hardware?
Will we incur any downtime during startup?
Is the solution flexible enough for all equipment on the floor? 
What is my OEE?
Will this help reduce downtime?

From a Supervisor’s Perspective:
What is the current running state of the equipment?
How do operators’ actions impact production?
How will an operator’s shift targets be displayed?
Will this help increase mold production?

From a Controls Perspective:
What known issues can be visualized to better understand how to correct them?
What cycle time improvements can be made without impacting quality?
How are similar machines running compared to each other?
Can we document and measure process tolerances?

From Maintenance’s Perspective:
What motions are trending negatively?
What are the occurrence and accumulation rates of fault messages?
Can information be automatically generated for maintenance teams?


Focusing on the questions developed during the success definition phase, Dotson Iron Castings and Neenah Foundry were able to clearly report out and determine if they had a successful adoption of analytics in their processes.

The executive level was able to answer all the questions posed with the solution chosen. Implementation was done remotely without any costly onsite shutdowns of machines. No modifications to equipment wiring or hardware was necessary. The only alterations were done with additional proprietary PLC logic embedded in the existing PLCs. 

Uptake or acceptance time is also vital to the implementation timeline. Dotson and Neenah ensured the tools selected were easy and intuitive to navigate and did not take lengthy and intensive training sessions. Employees from all levels must be able to use the data being gathered without interfering with their already full schedules.

A wide range of devices from various vendors were brought online and into the age of analytics––from brand-new equipment to 20-year-old machines with Ethernet capabilities.

The foundries saw:

  • A 15% reduction in maintenance costs compared to the last five years.
  • An 18% reduction in maintenance downtime in the targeted areas.
  • 10% improvement in mold 
  • production.

The supervisor levels can use the real-time data collection to quickly identify the running state of the equipment on the floor (Fig. 3). This reduced the reliance on line-of-sight inspection.

With the ability to view operator performance across shifts and multiple jobs, the supervisors can now identify those operators with optimal efficiencies and then train other operators based on setup procedures and task prioritization from high-performing team members (Fig. 4).

Line-side monitors were setup to display shift targets for the operators. A five-mold-per-shift increase was seen solely by giving the operators a constant update of where they stand against their targets. Shift leads set a target at the beginning of the work schedule and as the machine makes molds, a counter with color coding increases to easily identify accomplishments (Fig. 5).

In one instance, a machine was not making rate and needed to be sped up. Controls needed to ensure the cycle time reduction did not influence the quality of the molds being made. By visualizing the machine cycle, the team was able to reduce the cycle time for specific motions, which led to an overall cycle time reduction while still producing quality molds (Fig. 6).

One of the greatest benefits to collecting cycle time history from every motion or event is the ability to compare similar assets. This allows the capture of the best operating conditions across machines and applying them to every machine to set a standard tolerance and keep all machines running at an optimal level. Quickly having the tools to compare like jobs across similar machines has resulted in increased production.

The right data can ease the burden on maintenance teams. By providing insights to how machines are functioning, technicians can target which areas of machines to perform maintenance on, rather than relying on a static calendar or standard maintenance schedule. This reduces costs and downtime. 
By looking at the data proactively, an impending failure was caught before any downtime occurred. A warning trigger identified that a “Sand Gate Close” operation was trending negatively and taking longer than normal. The maintenance crew was alerted and preventive maintenance occurred during regularly scheduled maintenance. 

Another area that benefits from data collection is when maintenance is trying to conduct a forensic analysis after a failure has occurred. Active use of data can obviously catch problems like in the “Sand Gate Close” issue, but additionally, when a machine is allowed to run to failure, looking back at the data can help the team identify the issue that caused the failure, see when a failure was likely to occur, and understand when the failure was likely to have started its negative progression (Fig. 7).  

Maintenance teams also often battle phantom setting changes. Capturing all data and providing quick access to graphs and reports ensures that machines speed settings changes aren’t lost and can easily be viewed.

A problem machine is easily highlighted when plotting its fault occurrences over time. Often these alarms and faults can be silenced by operators, while never alerting maintenance of their presence. This keeps maintenance teams in the dark and sets them behind the curve when trying to keep their machines running smoothly.   

This article is based on the paper “Beyond the Buzz: A Practical Implementation of Industry 4.0 in Foundries,” (Paper 21-066) originally presented at the 2021 Metalcasting Congress.

Click here to read this article in the digital edition of April 2021 Modern Casting.