It’s great that we’ve sparked your interest in the topic of digitalization in industry! In our article, we highlighted why OPC UA unfortunately does not solve all the problems related to plant connectivity. Even if all plants are already equipped with OPC UA, MQTT and the like, we have had to learn the following.

Overview of the Blog Series on Connectivity & Machine Data:

1. OPC UA: Blessing or Curse for Industry 4.0?

2. Digitalization Dilemma: Working for the Data or Working with the Data

3. From Euromap, Data Blocks and Harmonized Data

4. Data Harmonization, or: What is My Throughput?

5. How Edge Devices Do Not Become a Security Gap

6. No More Closed Systems

7. Ready for All Challenges in Production with ENLYZE and Grafana

8. The Key to AI in Production

Lesson #2: Plant Connectivity is Not Even Half the Battle

Just because plants allow data to be queried does not create any added value for the company. This only arises when the data is used and insights are gained that lead to improvements in business processes or production. 

We currently see the challenges in four key points:

1. Integration of existing data sources from heterogeneous plant parks

2. Reading inconsistent and unstructured data formats

3. Unlocking and merging data silos*

4. Incorporation of distributed legacy software systems

*both time series data from sensors and booking data/orders

We observe three common approaches in the industry to address these challenges, which we will discuss in the following paragraphs. Finally, we will present how modern IIoT platforms attempt to solve the problems of current approaches and accelerate the path to Industry 4.0. 

🍝 Patchwork: Many Cooks, No Central Responsibility

The configuration we call Patchwork is the most commonly encountered solution in the industry. It resembles a patchwork quilt, as it has often emerged unplanned and has grown over time with the machinery. Various providers offer software that reads data from certain machines. For example, in an extrusion company, it looks like this: 

There is a system for material supply and monitoring of silo filling levels, as well as software for the main manufacturing step (e.g., extrusion systems) of the prevailing equipment manufacturer – which does not cover all machines or all parts of the so-called "Frankenstein plants". An additional software is responsible for post-processing (such as packaging and printing), and some data points are captured via a BDE or an MES.

The advantages of this solution are limited and primarily consist of the fact that they currently work. The disadvantages are clearly in the restricted usability and the work in data silos. Process engineers, developers, planners, and controllers would need to have multiple monitors on their desks as if they were monitoring a nuclear power plant, and are mainly busy exchanging data between systems via copy-and-paste. Furthermore, maintaining the various systems, coordinating suppliers, and continuing to develop functions is a significant effort.

Usually, the software was acquired through licenses, and upgrades are expensive and require lengthy projects. Required investments to modernize outdated systems are often so high that they are not profitable for the heterogeneous plant park.

Service is generally slow, and each provider focuses only on their own solution while interoperability between systems is neglected. The system resembles a tangled, spaghetti-like point-to-point connection. Keeping track and ensuring compatibility with updates or adjustments is a complex undertaking each time, which is often avoided.

🧰 In-house: Everything on Your Own

This configuration often occurs alongside the patchwork configuration, as individual solutions need to be linked through self-developed and managed systems. Installation companies commission components once and then leave the maintenance to the customer’s IT.

In other cases, companies take full responsibility for connecting and networking their plants and develop their own system for this purpose. When sufficient resources are available, this approach can be very successful, as the system is tailored to individual needs. Ideally, the solution is based on suitable technologies (software stack) and is not integrated into an existing ERP system. Since ERP systems are not suitable for processing large volumes of time series data, such integration often leads to a deterioration in overall performance. 

Wishful thinking: “Just take a Raspberry Pi and use it to read machine data” 

On the other hand, it can be problematic if the necessary resources are not available. An initial proof-of-concept (PoC) may be quickly implementable (often the simplest plant is selected), and the pilot operation usually runs smoothly. However, the complexity of maintenance and continued development is often underestimated. Particularly in the field of data management, the individual approach is difficult to scale. Device management, buffering solutions in case of internet connection failures, and redundant data storage are challenges that must be overcome once a pilot project is to be scaled.

🪤 Walled Garden: Trapped in One System

The Walled Garden approach is primarily pursued by hardware manufacturers who want to expand their business model through a software offering. This includes, for example, machine and control manufacturers. They offer a platform that is only compatible with their own machines or devices.

This approach is aligned with solutions like Apple’s App Store, which aims to bind customers to their own brand and thus promote the sales of their hardware. The success of such platforms shows why this approach is also attempted with IIoT platforms. However, implementation often fails because the platform primarily aims to create a lock-in effect, which many companies reject. Often, the solutions of plant manufacturers turn out to be extended HMI systems (“HMI with extension cord”) that do not optimally address the needs and questions of users.

We see one of the main reasons for the slow progress of Industry 4.0 since 2011 in the attempts of established players to develop such platforms. Instead of focusing on open systems where each integration increases the value of the product, plant builders and some traditional industrial software providers follow the usual path of developing everything themselves and only designing for their own products. Since interoperability is not pursued in this model, the customer ultimately suffers. As a result, many companies we talk to approach the digitalization systems of their plant or control builder with skepticism. After all, one does not want to commit to a single manufacturer. 

🪴 Vertically Integrated with Open Interfaces

The fourth approach includes both hardware and software components from a single source. In contrast to the Walled Garden model, however, the data in this system is not trapped but is accessible to other systems and providers.

To avoid the lack of central responsibility, as seen in the patchwork approach, it is important to provide a unified solution for a plant park with plants from various manufacturers. Since there are no conflicting interests, the focus is on the platform itself and the integration of as many data sources as possible.  This can usually only be achieved through a combination of hardware and software. 

These systems are not only manufacturer-independent from the perspective of plant connectivity, but also offer open, documented interfaces and integrations into third-party systems for easy and centralized data exchange. The focus is on systematically recording plant and production data, making it accessible, and enabling users to utilize this data in the applications of their choice. For example, PowerBI can be used for reporting, Grafana for monitoring, MS Teams for alarms, and Azure Data Lake as a data repository. Instead of settling for an inadequate all-in-one solution, users choose the best tool for their specific task.

Conclusion

The patchwork approach is widespread but often inefficient. In-house solutions can be successful if sufficient internal IT resources are available, but their complexity is often underestimated. Walled Garden models create dependencies on a single manufacturer, while vertically integrated solutions with open interfaces offer the best compromise.

Every company must build expertise in the field of digitalization and working with data to remain competitive in the future and attract qualified employees. The crucial question is in which areas to build competencies.

In our opinion, the focus should be as close to the company’s value creation as possible. For manufacturing companies, the emphasis should be on evaluating process data and building expertise in that area. The process experts are already within the company. When empowered to optimize the process based on data, the true ROI of a digitalization project will be achieved. 

Or in other words: Developing anomaly detection based on your deep process knowledge is more important than building scalable data pipelines that write thousands of data points per second to databases. Because: Working with data creates value for the company – Working for data is just a means to an end.

At this point, we would like to thank you for your renewed attention! As with the first article, if you have any questions or feedback after reading, please feel free to contact us at hello@enlyze.com. We look forward to hearing from you.