© shutterstock | Zapp2Photo



Nothing is changing value creation in the mechanical engineering industry as dramatically as the use of self-learning systems.

By Anke Henrich

Everyone is talking about artificial intelligence (AI) and in particular, the subsection of machine learning (ML). Although these technologies have their roots in the consumer trade, they are now growing in importance in the industrial sector and mechanical and plant engineering. Indeed, they are of existential importance, as software and informatic technologies are becoming a key driver of innovation. Cheap computing performance, as well as cloud and big data technologies have long since prepared the ground for this development.

Solving problems independently

One new factor is that, for the first time, computer programs based on machine learning can autonomously find solutions for new and unknown problems using algorithms. The artificial system identifies patterns and consistencies in the massive quantities of learning data it is provided with, while tools that are already established on the market provide the algorithms required for this. Thanks to AI, new and optimized software technologies, frameworks and platforms are entering the market almost incessantly.

As such, machine learning enables technical systems to do things that were previously the sole preserve of living beings: learning from experience and becoming more intelligent.

For example, ML algorithms learn what a good part looks like in a camera image using sample images. This then enables them to recognize and reject bad parts.

Although this appears to be a convincing argument, many mechanical engineering companies are still not sure whether machine learning is relevant to their business. This is also evident in the IT Report 2018, in which 27 percent of surveyed companies stated that ML still only plays a minor role. At the same time, 25 percent believe that AI is of moderate importance.

The business fundamentals in mechanical engineering are shifting, with some sectors developing more quickly than others. Due to this development, the increasing interchangeability of individual machines will mean that in some sectors demand will no longer focus solely on the machine itself, but also on additional services such as predictive maintenance.

Optimized products and processes

Machine learning allows existing business and production processes to be optimized. As a result, machines are maturing into intelligent process providers that work in a virtually autonomous way. Machine learning is not only enhancing product properties, however - it is also optimizing internal processes, such as those for incoming payments, quotations and production planning.



The characteristics of ML also differ in the products themselves. For example, the machine operator is supported by expert systems integrated in the machine. Meanwhile, AI provides benefits in the form of maintenance or additional services in the process environment of the machine. This pays off, too: For instance, an automated incoming payment process with invoice comparison can bring about savings in the double-digit percentage range.

Services with added value

Requests for quotations for complex machine configurations are also changing. Thanks to ML-based automation, these requests are answered much more quickly and the provider’s number of contract completions will grow accordingly. Scenarios like these are already a reality for ERP, marketing or sales systems. They reduce familiarization periods, training costs and set-up times while simultaneously boosting efficiency.

For customers, this provides immediate added value when operating the machines, while at the same time enabling providers to offer new added-value services based on the available machine data - for instance with predictive maintenance. Impending problems are detected and remedied thanks to early data analysis, thereby preventing downtimes - an enormous advantage for companies with closely timed work processes. ML also simplifies the tasks of customers when it comes to the documentation obligation.

On the basis of these and similar algorithms, mechanical engineering companies can develop new customer services as well as more differentiated pricing models for using the machines. The hope is that ML will lead to zero-error quality while also keeping services on time.
The technical possibilities are obvious, however, finding the right approach to the topic is not. Studies show that many companies make their first projects too large and complex while overlooking relevant aspects.

Therefore, a variety of important questions must be formulated and answered in detail before beginning the project. These should focus on relevance, risks and return on investment.

For example:

  • What data do we have?
  • Which application scenario is the most promising starting point for my company?
  • How can we build up the required knowledge in the company in a step-by-step manner?
  • Which ML techniques and algorithms are relevant for us?
  • What is the potential risk of the chosen algorithms for our company – and how can we counteract it at an early stage?

And last, but not least: Who in the company is responsible for delegating which decisions to machines?

With its experts from the mechanical engineering industry, automation specialists and software companies, VDMA Software and Digitalization will help ensure that your company’s path is a successful one.

The association will help you analyze opportunities, benefits and risks in a structured way, with practical examples setting the individual aspects in a business context.

Further Information

VDMAimpulse 01-2019 "The big give and take"   |   VDMA Software and Digitalization   |  VDMA IT-Report 2018-2020

Vanessa Koller, VDMA Software and Digitalization.