Harnessing the potential of machine data

The print shop is a hive of activity: thousands of brochures are printed, printing inks are refilled, paper is prepared for processing and printed products are dispatched. Artificial intelligence optimizes production by analyzing machine data.

The printed sheets come out of the machine 6 milliseconds earlier. For humans, this is less than the blink of an eye, but for printing machines, it is a magical threshold at which their transport wheels need to be cleaned. Only artificial intelligence registers this barely perceptible difference and can process this information. Only AI technologies can extract valuable information from machine log files, revolutionize maintenance and repair and usher in the age of industrial intelligence.

The AI-based machine monitoring system has already predicted that the paper transport wheels are slowly becoming dirty and will need to be cleaned after printing another 200,000 sheets at the latest. The production manager is prepared for this and plans his production accordingly in order to complete the order on time.

Industrial analytics optimizes more than just machine availability

What looks like a normal day in a print shop is the next stage in the development of Industry 4.0: industrial analytics. A discipline that has set itself the task of harnessing the immense potential of machine data to optimize maintenance and production processes.

Analyzing machine and process data is nothing new and has been common practice for critical systems for decades. However, appropriate measures are often only initiated reactively, as the logs are first read out and then analyzed offline. In the worst case, a need for action is only recognized when a serious problem already exists. In this way, it is not possible to continuously monitor a global machine fleet and take proactive action.

This has changed fundamentally thanks to today's AI processes: Sensor, log, process and production data are recorded by digital, automated analysis processes and processed in the cloud with high performance. In this way, information about the plant can be evaluated in real time.

Such an approach is scalable and an important prerequisite when it comes to providing predictive service for several thousand machines or systems worldwide. However, the new options for evaluating the database are only the first step in gaining insights for production, maintenance and further development of the products.

From asset to action - and back again

In addition to the right algorithms and high-performance analysis, it is just as important to manage the use of information and findings within the company. The next step towards Industry 4.0 therefore requires system integration in order to proactively initiate maintenance measures thanks to the predictions or to use new findings to continuously optimize machine performance. By connecting ticketing systems, for example, new tickets can be created automatically as soon as critical values are exceeded. Similarly, thanks to modern AI processes, the corresponding assembly and maintenance instructions can be sent directly to the service technician, depending on the affected component.

System integration is not a one-way street, as it must also be ensured that information about measures actually carried out and knowledge gained in the process is continuously fed back into the case analysis, creating a cycle of learning and improvement.

The questions go into detail:

• In how many cases was the prediction correct?

• Which further analyses of log data are relevant?

• Which product type shows the errors most frequently?

• Could there be a serial error?

Digital transformation in service

However, the development of such integrated systems in the manufacturing industry is not an end in itself and goes far beyond the goal of higher machine availability. In addition to maintenance, the potential of data for production optimization is often overlooked by many companies.

Instead, they work manually with lists and Excel spreadsheets and rely on the knowledge of individual experts - a risky undertaking in the face of demographic change. It's a shame really, because so much more could be done with data analysis and new revenue could be generated, for example by offering maintenance planning based on machine data as a new service that proactively prevents breakdowns and offers the right spare parts at the same time.

Suppliers and customers alike benefit from such models: Improved operation secures production capacities, while information on the maintenance and repair measures actually carried out makes the system increasingly precise in its predictions.

In addition, insights into the overall equipment effectiveness (OEE) can be gained from the operator's perspective. In order to make optimizations here, however, the entire process must be analyzed coherently with regard to the three core factors of plant availability, productivity and quality and constantly adapted to changing external influencing factors, such as fluctuating raw material quality. Due to the high complexity and volume of data, it is time to expand the experience of individual employees with AI-based data analysis in order to optimize processes. This is the only way to achieve objectification and systematic improvement.

Industrial analytics therefore focuses on the business value of the data, not just on the analysis itself. It's not about the algorithms, but about operators understanding their machines, devices and processes: How do they really work? How can they be made measurable? And how can their behavior be predicted? Industrial analytics therefore accompanies the entire process: from data analysis to the right measures. This enables companies to optimize their spare parts and operating resources business, develop new data-driven business models and expand their range of services.