A machine and plant manufacturer wants to use artificial intelligence (AI) for itself. How does he make the right choice from the multitude of options?

Markus Günther is Product Manager in the Production division at Inform GmbH. He is responsible for the Felios APS system. © Inform

Every company is individual and faces its own unique requirements. Nevertheless, it makes sense to start with a standardized software solution. Over time, this can then be adapted to the specific needs of your own company or expanded accordingly. To do this, project managers can take a look at what established providers already offer on the market. As a rule, it is also worth taking a look at what other companies are already implementing in this direction.

In terms of timing, the weakening order situation offers a good opportunity to address the issue. Manufacturing companies can see the phase of a quieter economy as an opportunity to dedicate themselves to such digitalization projects and emerge stronger.

How do manufacturing companies choose the right solution provider for AI projects?

When choosing a solution provider, companies should look for a balanced combination of industry expertise and development skills. On the one hand, it is important to know the specific needs of the user companies, while on the other hand, technical know-how is required with regard to the algorithms.

What does machine learning actually mean for machine and plant manufacturers?

First of all, it is important to understand that ML does not just make machines smarter. Many people immediately think of classic use cases such as predictive maintenance when they hear the term - but this is only a small part of the overall picture. ML can also be used to optimize processes and services. For example, ML-supported systems can use historical data to make reliable predictions about developments, such as the replenishment time for purchased parts or the duration of certain work processes. ML thus enables more precise planning and improves both process quality and adherence to deadlines. This gives machine and plant manufacturers a decisive competitive edge in a highly competitive market. A major advantage is that the data required for this is usually already available in the company, for example in the context of an ERP system.

Which existing company data can generally be used for ML?

Which data can be used for the ML application also depends on the company in question. Companies are often surprised at the valuable data they already have at their disposal. For example, order data from previous years is stored in the ERP system. An ML algorithm can use this data to forecast future delivery dates or replenishment times and thus enable more precise production planning.

In general, companies often think too much about their data quality. In many cases, relatively coarse data is sufficient, and some data is not even needed - for example, the buffer times between individual processing steps. Manufacturing companies should therefore not allow such concerns to deter them from important digitization projects, but should rather tackle the matter with the support of a service provider.

What role do ERP systems play in connection with AI technology and when and where do they reach their limits?

ERP systems are ideal for collecting and managing data. With this in mind, they are essential for the efficient coordination of complex production processes. However, they do not contribute to extracting increased added value from the available data. They therefore reach their limits at the latest when it comes to evaluating, let alone optimizing, data. AI-supported systems then take the next step: they build on the ERP system and compensate for its planning weaknesses through precise production control.