The manufacturing industry, like so many other sectors, is being shaken up by the disruptive power of modern technologies. New concepts - such as the Industrial Internet of Things (IIoT) - enable companies to quickly ramp up and ramp down their production capacity as needed, shorten production times and minimize downtime. They can also identify and improve complex interrelationships in the operating process much more easily thanks to the detailed monitoring of systems. However, for IoT to be used successfully in production plants, there are a few points to consider: laying the foundations and creating the right infrastructure as well as visualizing, analysing and ultimately optimizing the entire plant.

Laying the foundations

First of all, plant operators need to determine which aspects they want to focus on in order to optimize production with the help of IIoT: For example, should output, the quality of the units produced, the maintenance cycle or resource consumption be monitored?

As the key figures for optimum operation or the maintenance cycle vary from machine to machine, an inventory is necessary. The data required for the analysis is not available for all systems. Plant operators can retrofit some machines and avoid having to purchase a new device. It is worth asking the manufacturer whether the required key figures will be available through retrofitting.

In addition to the (real-time) data from the sensors, structured and unstructured data from control systems, logs, databases, SCADA systems or similar sources must also be collected for analysis in order to gain an accurate overview of production.

The right infrastructure

Parallel to selecting the data to be analyzed, the infrastructure should be prepared. Most industrial processes operate at very high speeds and therefore generate extremely large volumes of data. Processing this data in a central cloud, for example, is an unnecessary detour and not economically advisable. In addition, the cloud does not always offer 100% protection against access by third parties; constant bandwidth - which is essential for such an IT operating model - is also not always guaranteed.

This is why local analysis of the data near the machine - i.e. at the edge of the network - is required. Processing is significantly faster with this method, it is also more secure and consumes significantly less bandwidth. However, the cloud can be used to store data that has already been processed.

Visualize and analyze

The processed data should then be visualized in a meaningful way. The type of visualization depends on the type of machine, the data generated and the operating parameters. For example, repetitive patterns can be visualized and the performance of the machines can be read more easily. This saves time when optimizing production processes.

Such analyses can also be used to identify complex correlations that depend, for example, on the number of units produced or the outdoor conditions during production and can affect the condition of the machines or the quality of the processes and products. If, for example, a change in the outside temperature frequently leads to errors or failures in production, plant operators can counteract this with rules that have been defined in the plant's operating process.

A suitable IIoT solution can then automatically monitor regular operation and run a defined protocol in the event of deviations.

How the system becomes really smart

If you want to build a smart factory that is not just "connected" and collects data, you will benefit from the use of machine learning (ML). Using ML, machines can use the collected data to learn how to adapt to external conditions, for example, without having to be reprogrammed. An intelligent thermostat on a machine could then analyze data - such as indoor or outdoor temperature, time of day and season - and learn to predict the likelihood of a production disruption due to overheating. In a further step, the thermostat could automatically reduce the speed to prevent a breakdown.

ML can also be used for the continuous improvement of defined control procedures and for the detection of anomalies. In contrast to pre-defined rules, in such a scenario the system can react to other factors that have been learned over time and were not yet included in the programming. For example, the system could detect minimal deviations in regular operation and send an SMS or email before a failure occurs. Depending on how serious the deviation is, the system can either notify the staff or possibly order the necessary spare parts itself.

A truly smart production plant can thus save on resource consumption and optimize processes during production and the maintenance cycle.

The author

Dorian Guest, Dell © Dell

Dorian Gast is Head of Business Development IoT Germany, Israel, UAE at Dell Technologies