Artificial Intelligence in the Machinery Industry
In several publications, drupa’s supporting association VDMA is taking a close look at artificial intelligence for the machinery industry. The potential AI carries to support the machinery industry is very high, thus we are examining the steps that are needed to successfully incorporate AI and also giving details about examples of how it could be used.
In order to assess the potential of digital transformation for companies in the mechanical and plant engineering industry and thus also for important players in the printing industry, the supporting association of drupa VDMA, the largest network organisation and an important voice for the mechanical engineering industry in Germany and Europe, has taken a closer look at artificial intelligence in the machinery industry in their online magazine “Digital Transformation – Digitisation changes the world” (German only).
With artificial intelligence, a technology is available in which huge amounts of data can be structured, analysed and evaluated. The areas of application are manifold: intelligent recommendations, process improvements, anomaly and error detection including root cause analysis, predictive maintenance optimisation and the type and speed with which products are designed and produced.
The Four Steps of AI
There are four basic steps that need to be taken in most cases when it comes to artificial intelligence to reach the desired result. In his article “With AI, analytics and smart data a step ahead” (German only), Wolfgang Ennikl, CEO at Cubido Business Solutions, elaborates on these, starting with the first one: a definition workshop.
Here the use case is defined and a way is outlined how the goals of the use cases can be achieved. Particularly important in the field of AI is on the one hand the definition of the task that the AI has to solve and on the other hand the data that the AI needs to solve the task in a satisfactory manner. Both points must be clearly defined at the beginning of the project.
After defining the necessary data basis, the data is recorded in the second step. This data may be enriched with labels. In this step the basis for AI is created, so it is of particular importance to pay attention to high data quality.
Then follows the third step, in which the collected data and the knowledge contained therein are poured into models. However, the training of the models also includes an evaluation strategy, because only through a precise evaluation can we be sure that we have defined the task that we have given the AI precisely enough and have collected the right data.
Lastly, the tested models can now be used productively. However, this is not the end of the project, because further data can continuously improve the model. Developing AI is an iterative process that works closely with the process in which the AI is involved in. AI can improve the process, which in turn entails new tasks for AI and thus requires new data.
Usage of AI: Examples
The question that remains now is: What exactly can AI in the machinery industry be used for after having completed these steps? The first example is “automatic optimisation of setting parameters”. With the use of machine learning, work steps in the production process of companies can be shortened while at the same time increasing quality by up to 75%. It optimises the workflow. AI can also be used in the area of risk assessment, for example to determine environmental influences on the composition of agricultural products as well as the prognosis of the risk of fungal infestation.
Furthermore, AI is able to successfully identify influencing factors on product quality. Data is used as a basis and with further analyses it can be determined which processes can be significantly improved. Additionally, influencing factors on product quality can be identified. This ensures higher quality at lower costs. Lastly, AI is also helpful in the area of quality assurance and production process optimisation. In extensive production processes, such as in a joint project with several manufacturing industrial companies, a huge number of influencing factors play together to form the finished product. However, checking for errors is only possible at the very end of the process. Therefore, in such cases, a common way must be found to identify harmful production conditions even before waste is produced. In addition to the correlation of data of the most diverse origins, the correct transformation of sensor data is a particular challenge in such cases. With the help of statistical analyses, changes in the use of raw materials or adjustments to process parameters can be derived from the generated data. By optimising these input variables, rejects can be reduced and thus also cost savings.
The potential AI offers and the utility it already provides make it a very valuable resource in the world of the machinery and printing industry.
Could you imagine other uses for AI in the printing area? If so, which one? Tell us in the comments!